<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by ISA Manipal on Medium]]></title>
        <description><![CDATA[Stories by ISA Manipal on Medium]]></description>
        <link>https://medium.com/@isa-manipal?source=rss-e1741a63268c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*Httsij69u4PdszsNLMCx5A.png</url>
            <title>Stories by ISA Manipal on Medium</title>
            <link>https://medium.com/@isa-manipal?source=rss-e1741a63268c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 01:32:05 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@isa-manipal/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Reinforcement Learning.]]></title>
            <link>https://isa-manipal.medium.com/reinforcement-learning-5ad7971a08f5?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/5ad7971a08f5</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[reinforcement-learning]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Thu, 14 Oct 2021 08:32:47 GMT</pubDate>
            <atom:updated>2021-10-14T08:41:30.807Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Research and writing by Yashwanth — ISA, Manipal.</blockquote><h3><strong>What is Reinforcement Learning</strong></h3><p>Reinforcement learning is a machine learning training method. It consists of an Agent (something that can perceive its environment and make decisions), the actual environment, and an interpreter.</p><p>Reinforcement learning is used to train a model to make a sequence of actions. This is useful when we want to train models to surpass human-level competency, or the action is too complex to be described by a human. For example, how can we explain to another person how to walk? As you can see, it is an incredibly challenging task. This is where Reinforcement learning comes into the picture.</p><p>Learning starts with random decisions made by the agent using trial and error. When the goal is achieved, the interpreter rewards the agent, and when the goal is not achieved (or achieved incorrectly), the agent is punished. This way, the agent makes the best decisions to reap the maximum reward and hence, “learns” how to do the task efficiently. The “state” is the current situation of the agent and it creates the context necessary for the agent to make the next decision in the sequence.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/250/0*PjUcprMmEN0gCZzr" /></figure><p><strong>Applications</strong></p><p>Reinforcement has become significantly more popular after the introduction of modern computing power. Some of the popular applications for reinforcement learning are:</p><p>Robotics: Most self-driving cars or autonomous machinery use some aspects of reinforcement learning. For example, in self-driving cars, the model is given a video feed and other data. It is then trained to drive on the road safely while going at an optimal speed. It needs to consider speed, proximity to other cars, traffic, road type, the safety of the driver, and many other factors. The challenges of training such a model are clear.</p><p>Games: Reinforcement learning is also used in video games, commonly in the forms of Q-learning and policy search. It is used in pathfinding, NPC(Non-Player Character) actions, and the making of AI-controlled opponents.</p><p>Q-learning is a model-independent reinforcement learning algorithm that is used to determine the value of any particular action.</p><p>Policy search is a systematic approach to embed expert knowledge at the initial stages of the model training. Essentially, it is the process of choosing good parameters to start with to ensure faster training and higher accuracy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/623/0*HkYF3XwFBi5up23n" /></figure><p>Bio-mechanics: Biological functions are difficult to simulate and teach to machines. Reinforcement learning allows the learning of complex biological actions, such as running. The “Learning to run” project conducted in Stanford’s Neuromuscular Biomechanics Lab is a great example. Reinforcement learning is also used in the simulation of new prosthetics.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/471/0*SDp9n7TVWL-XPyop" /></figure><p><strong>Challenges</strong></p><p>Environment: An accurate simulation environment is very challenging to make. Even if the model performs extremely well in the environment, there is no guarantee that it will be just as effective in the real world. This is a large issue in self-driving cars due to the need for high reliability and high cost of failure.</p><p>Credit assignment problem: Traditionally, reinforcement learning models only reward the agent if it completes a task successfully, hence disregarding all the steps leading up to it. Even if most of the steps are correct, the interpreter gives the agent the same reward/punishment as if it had failed completely. Hence, each action that leads to a higher cumulative reward needs to be given more value, or “credit”. Deciding the value of each action is also known as reward shaping.</p><p>Reward over goal: Often, you will see models using reinforcement learning finding loopholes to get more rewards, but not achieving the desired goal. For example, a car continually picking up coins in a game without finishing the race. This is part of the reason reinforcement learning is challenging to implement.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/478/0*psgg9HyTFTb6N6Rl" /></figure><p><strong>The way forward</strong></p><p>Reinforcement learning is extremely important and can be applied to solve many complex issues. We observe that the media often portrays reinforcement learning applications as something out of a science fiction movie. In reality, these complex problems have been solved due to the dedication and hard work of the brightest minds in engineering. It is important to know the limitations of reinforcement learning today and work to further explore the field.</p><p>One day, we might reach a point where we can deploy completely autonomous robots. Of course, technology is moving much faster than laws, and it is important to create and enforce the necessary legislation to protect people.</p><p><strong>References</strong></p><ol><li><a href="https://deepsense.ai/what-is-reinforcement-learning-the-complete-guide/">https://deepsense.ai/what-is-reinforcement-learning-the-complete-guide/</a></li></ol><p>2. An Introduction to Reinforcement learning by Arxiv Insights.</p><p><a href="https://deepsense.ai/learning-to-run-an-example-of-reinforcement-learning/">3. https://deepsense.ai/learning-to-run-an-example-of-reinforcement-learning/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5ad7971a08f5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Zero Knowledge Proofs and their applications in IoT]]></title>
            <link>https://isa-manipal.medium.com/zero-knowledge-proofs-and-their-applications-in-iot-4443e5025fd8?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/4443e5025fd8</guid>
            <category><![CDATA[iot]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Thu, 14 Oct 2021 08:10:49 GMT</pubDate>
            <atom:updated>2021-10-14T08:10:49.795Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Introduction:</strong><br>Zero knowledge proofs are cryptographical methods by which one can prove to another that they know a certain value “z” without conveying the actual value “z” to the other person. Understanding them will be easier with the help of an example —</p><ul><li>Alice is color-blind and Bob isn’t</li><li>Alice has two balls, one green and one red. She is unable to tell the difference between the two.</li><li>Alice does not believe that Bob is not color blind</li><li>To verify his claim, Alice holds both balls in her hands and tells Bob that she will either switch/not switch them behind her back, but Bob must tell her if she has switched the balls. If Bob is not color blind, he will be able to tell whether she has switched the balls.</li><li>In the first try of Bob guessing if the ball has switched, he has a 50% chance of guessing it right. Hence, Alice conducts the experiment 20 times in a row. The probability that Bob guessed all of them are right is 9.53674316e^-7. Which is sufficiently small to assume that Bob is, in fact, not colorblind.</li><li>This way, one can confirm the knowledge of the prover without the verifier knowing anything about it.</li></ul><p><strong>History:</strong></p><ol><li>Zero Knowledge Proofs were conceived by 1985 by <a href="https://amturing.acm.org/award_winners/goldwasser_8627889.cfm">Shafi Goldwasser</a>, <a href="https://amturing.acm.org/award_winners/micali_9954407.cfm">Silvio Micali</a>, and <a href="http://www.cs.toronto.edu/~rackoff/">Charles Rackoff </a>in their paper “<a href="http://scholar.google.co.in/scholar_url?url=https://core.ac.uk/download/pdf/194164868.pdf&amp;hl=en&amp;sa=X&amp;ei=QYRxYJ6SFceSywS5zYnwBg&amp;scisig=AAGBfm3iP2tEPqqwdmm6qMl_rddbqZR8Mg&amp;nossl=1&amp;oi=scholarr">The Knowledge Complexity of Interactive Proof-Systems</a>”.</li><li>Goldwasser, Micali and Rackoff wrote mainly about Interactive Proofs(IP)</li><li>In 1988, Non-Interactive Zero Knowledge Proofs were conceived by Blum, Feldman, and Micali. They are also known as zk-SNARKs/zk-STARKs.</li><li>A common reference string shared between the prover and the verifier is sufficient to achieve computational zero-knowledge without requiring interaction. Goldreich and Oren gave impossibility results for one shot zero-knowledge protocols in the standard model. In 2003, Goldwasser and Kalai published an instance of an identification scheme for which any hash function will yield an insecure digital signature scheme.</li><li>The first widespread application of zk-SNARKs was in the <a href="https://z.cash/technology/">Zerocash blockchain</a> protocol, where zero-knowledge cryptography provides the computational backbone.</li><li>It is assumed that By this year, 82 <a href="https://en.wikipedia.org/wiki/Cryptocurrency">cryptocurrencies</a> worth a total of US$8.85 billion encrypted their transactions with zero-knowledge proofs or a similar private technology.</li></ol><p><strong>Precursory Information:</strong></p><p>To understand the math behind it, one must have knowledge of Elliptic Curve Cryptography, or ECC in short.</p><p>An elliptic curve is the set of points that satisfy a specific mathematical equation. The equation for an elliptic curve looks something like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/194/1*NIPLmrs1BVFdOUt_mOq8Lg.png" /><figcaption>Example Equation</figcaption></figure><p>The graph looks like this:</p><figure><img alt="ecc&#39;s trapdoor function example" src="https://cdn-images-1.medium.com/proxy/1*ZrzlOZDcYIFVwpypm4w7TQ.gif" /></figure><p>Elliptic Curves have symmetry about the x axis, and any non-vertical line will intersect the curve in at most three places.</p><p>First, we start with an arbitrary point on the curve. Next, we use the dot function to find a new point. We keep repeating the dot function recursively.</p><p>If you do this 3 times you may notice that you end at a point E. If you know the starting point, ending point and number of hops it is easy to land at that point. However, if you have no knowledge of the number of hops, it is nearly impossible to arrive at the ending point.</p><p>Practically, ECC relies on the existence of a private key and a public key. The private key is used to encrypt the data into a hash, and that hash can be used to verify the owner.</p><p>Think of the public key as the starting point and ending point, and the private key as the number of hops.</p><p>ECC is important because it provides the same level of encryption as RSA, but with a much reduced bit count.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/431/1*RrwkUgYeGZF3GnkF3rmu9g.png" /><figcaption>Comparison between RSA and ECC decryption time</figcaption></figure><p>Quadratic Residue: In number theory, an integer q is called a quadratic residue modulo n if it is congruent to a perfect square modulo n; i.e., if there exists an integer x such that:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/178/1*NhbKASJx4eubOE3raCur4w.png" /><figcaption>Quadratic Residue</figcaption></figure><p>A great explanation for quadratic residues is provided by <a href="https://www.youtube.com/watch?v=34yBBgd_Jpk">Polar Pi</a>.</p><p><strong>The Math:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/776/1*_Fgowpkx2Xr0eBQNsaVZsg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/746/1*7Q5PK_mHVRMeuYNNkii1Zw.png" /></figure><p><strong>zk-Proofs and IoT:</strong></p><p>Internet of Things (IoT) devices are being deployed in huge numbers around the world, and often present serious vulnerabilities. Accordingly, delivering regular software updates is critical to secure IoT devices.</p><p>Manufactures face two predominant challenges in providing software updates to IoT devices:</p><p>1) scalability of the current client-server model and</p><p>2) integrity of the distributed updates — exacerbated due to the devices’ computing power and lightweight cryptographic primitives</p><p><em>Crowdpatching, </em>is a protocol leveraging zk-proofs and Blockchains to deliver security patches to IoT devices.</p><p>In this system, the manufacturer delegates the delivery of new updates to self-interested agents, called distributors, who undertake this task in exchange for cryptocurrency (CC) payments.</p><p>In this context, the use of a zero-knowledge tool, called zk-SNARKs, is fundamentally important.</p><p>Before a proof-of-delivery can be issued, a distributor is expected to provide a zk-SNARKs proof about the provided encrypted update, mathematically proving that: 1) it was indeed obtained encrypting the update file authorized by the manufacturer and 2) the employed key has indeed the provided hash value.</p><p>Most importantly, this zero-knowledge proof does not reveal anything else beyond those facts: the unencrypted update and the key remain secret.</p><p>If the proof is valid, then the proof-of-delivery signature is generated by the IoT device, and sent to the <a href="https://ethereum.org/en/glossary/#smart-contract">Smart Contract</a> by the distributor along with the encryption key.</p><p>Before issuing the payment, the contract checks that the signature is valid, and that the key matches the hash value. Along with the payment, the SC publishes the key on the BC. The latter can be now used to decrypt the file, and the update is finally obtained by the IoT device.</p><p>This was one example of many, on how the strong cryptography protocol of zk-SNARK’s can secure IoT devices.</p><p><strong>Limitations of zk-Proofs:</strong></p><ul><li>The computational power required to generate a proof is significantly higher than other protocols.</li><li>The space taken up by said proofs are significantly higher than other protocols as well.</li><li>If the prover loses access to the secret, it is lost forever (this could be an advantage in some cases)</li></ul><p><strong>Current Applications:</strong></p><ul><li>Zero Knowledge Proofs are used widely in truly private cryptocurrency transactions, as seen on the prominent L2 solution, <strong>ZK-Rollups</strong></li><li><em>Crowdpatching</em></li><li>Highly secure authentication systems</li><li>Securing Intellectual Property</li></ul><p><strong>References:</strong></p><ul><li><a href="https://crypto.stanford.edu/pbc/notes/crypto/zk.html">https://crypto.stanford.edu/pbc/notes/crypto/zk.html</a></li><li><a href="https://eprint.iacr.org/2019/099.pdf">https://eprint.iacr.org/2019/099.pdf</a></li><li><a href="https://core.ac.uk/download/pdf/194164868.pdf">https://core.ac.uk/download/pdf/194164868.pdf</a></li><li><a href="https://www.youtube.com/watch?v=34yBBgd_Jpk">https://www.youtube.com/watch?v=34yBBgd_Jpk</a></li><li><a href="https://arxiv.org/pdf/2010.12134.pdf">https://arxiv.org/pdf/2010.12134.pdf</a></li><li><a href="https://www.cs.umd.edu/~jkatz/papers/bbZK.pdf">https://www.cs.umd.edu/~jkatz/papers/bbZK.pdf</a></li><li><a href="https://github.com/scipr-lab/libsnark">https://github.com/scipr-lab/libsnark</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4443e5025fd8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Active Roll Over Protection System]]></title>
            <link>https://isa-manipal.medium.com/active-roll-over-protection-system-abbdd48f5c9b?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/abbdd48f5c9b</guid>
            <category><![CDATA[automobile]]></category>
            <category><![CDATA[mechanical-engineering]]></category>
            <category><![CDATA[cars]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Thu, 07 Oct 2021 13:29:36 GMT</pubDate>
            <atom:updated>2021-10-07T13:29:36.289Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Research and writing by Mechanical sub-system, ISA-Manipal student section.</blockquote><p>It has been almost 200 years since first commercial vehicle has been in our world. We have been making progress ever since, improving speed, efficiency, quality of life and much more. Travel has become a necessary part of our life and many people’s lives are dependent on it. Be it heavy duty transportation by trucks or daily commute through city roads, more and more people are integrating travel in their lives and hence safety becomes an important metric of concern. According to WHO approximately 1.3 million people die due to accidents regarding travelling by any vehicle. It maybe speeding, distracted driving, unsafe roads, vehicles, etc. Speeding accounts for more than 60% of Accidents and hence is a major area of importance.</p><p>Over speeding has been a major problem recently as the world moves to more strict timing and deadlines, drivers try to compensate by speeding. Vehicle becomes very hard to control during uncertain situations. Active Roll over protection system helps drives to better control their vehicles by using various Sub-mechanisms.</p><p><strong>How it works?</strong></p><p>ARP for heavy commercial vehicles is basically an electronic control system built upon the following main sub-components — Anti lock Braking system, Yaw and Traction control. ARP can also detect forthcoming rollovers (rollovers is a type of crash in which vehicle is tripped over to its side or roof) during speeding over a sharp turn and can also prevent rollovers using suspension system to counteract the center of gravity of the vehicle.</p><p>Anti-Lock Braking System — This system prevents locking to brake plate-wheel which is necessary because when the wheel locks in and at high speeds the vehicle skids on the road making it harder to control and deceleration decreases.</p><p>This system works by actively checking to rotation speed of wheels and correcting the brake pressure in case of rotation of wheel doesn’t correspond to speed of vehicle.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/330/1*B39dMMpwA4lxw3obinjksQ.png" /><figcaption>Speed sensors that can be attached to wheels axle</figcaption></figure><p>It consists of electronic speed sensors for each wheel and at least two hydraulic valves within brake system. Speed sensors act as a control signal to hydraulic valves which can check rotation speed up to 15–30 times second in modern vehicles. After detecting mismatched conditions, it makes necessary changes. The ratio of changes is generally pre-defined based on vehicle use case after trying and testing on different roads, whether conditions, Tire material used, etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/623/1*q7vcnfRH4p8EeunIQNqTXQ.png" /></figure><p>Modern stability control systems are complex systems based on the ABS. Here, a minimum of two or more sensors are added to help the system to be functional. A Gyroscopic sensory and Steering wheel angle sensor. During sharp turns gyroscopic sensors and steering wheel sensors should be in sync otherwise the vehicle wheel would lock itself and skid, hence when they are in a mismatch a software makes necessary adjustments to brake different wheels to produce the same effect as intentional turn made by the driver.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*KPX8WFkpZDoH3dsfzqTwNQ.png" /></figure><p>Traction Control System — This system prevents the wheel from slipping on acceleration or when driving on smooth and slippery surfaces. Traction control takes the input of the speed from each wheel using speed sensors. The onboard traction control system measures the difference in rotational speed of each tire and hence, determine which wheel has lost traction. If the system detects a wheel is slipping, the brake module applies pressure to the brakes of that wheel to reduce its speed. More advanced traction control systems can also reduce the engine power going to the wheel that is slipping and transfer it to a wheel that have more grip.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/266/1*xNPrK9XO3EnoDbJiU9qL9w.png" /></figure><p>Yaw Control System — This system sustains longitudinal acceleration and improves lateral stability by controlling torque distribution between the driven wheels. It is generally used to maximize traction while cornering on sharp turns. The torque distribution is controlled by the ECU and is applied to the wheels through several mechanical components in the differential. In some cases, the clutch pressure is also varied to control the amount of torque going to the transmission system. If a vehicle begins to rotate in a manner inconsistent with the steering wheel position, torque from the outside wheel will be transferred to the inside wheel. As a result, the vehicle will rotate in the direction as inputted by the steering wheel.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/420/1*JD9cYcGdrELiKqHatM-mNg.png" /></figure><p>Conclusion:</p><p>As said earlier, the ARP system works with the help of the Anti-Lock Braking System, Traction Control System and the Yaw Control System. When the driver turns suddenly, inputs are taken in from the steering angle, relative speed of the wheels from the speed sensors and tilt from the gyroscopic sensor. The traction control system prevents the wheels from slipping, thus, ensuring the vehicle stays on its path. The Yaw control system will take the input of steering angle and speed, then accordingly, distribute the torque between the wheels (more to the outer wheels compared to the inner wheels). And finally, if the driver is applying brakes while performing a sudden manoeuvre, the ABS module will apply brake pressure in pulses so that the wheels don’t lock up and causes the tires to slip. The gyroscope sensor keeps in check how much the vehicle is leaning while turning and accordingly apply brakes or limit the torque.</p><p>In this way, the ARP system works with its sub-systems to ensure that the driver can perform sharp and sudden manoeuvres without losing control of his vehicle and ensuring the safety of the passengers onboard.</p><p>References:</p><p><a href="https://www.who.int/news-room/fact-sheets/detail/road-traffic-injuries">https://www.who.int/news-room/fact-sheets/detail/road-traffic-injuries</a></p><p><a href="https://www.monash.edu/__data/assets/pdf_file/0020/218270/racv-abs-braking-system-effectiveness.pdf">https://www.monash.edu/__data/assets/pdf_file/0020/218270/racv-abs-braking-system-effectiveness.pdf</a></p><p><a href="https://www.nhtsa.gov/risky-driving/speeding">https://www.nhtsa.gov/risky-driving/speeding</a></p><p><a href="https://cecas.clemson.edu/cvel/auto/systems/yaw-control.html">https://cecas.clemson.edu/cvel/auto/systems/yaw-control.html</a></p><p><a href="https://auto.howstuffworks.com/28000-traction-control-explained.htm">https://auto.howstuffworks.com/28000-traction-control-explained.htm</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=abbdd48f5c9b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Bomb Detection and Disposal Robots]]></title>
            <link>https://isa-manipal.medium.com/bomb-detection-and-disposal-robots-1c5035507b54?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/1c5035507b54</guid>
            <category><![CDATA[military]]></category>
            <category><![CDATA[army]]></category>
            <category><![CDATA[explosives-detection]]></category>
            <category><![CDATA[robots]]></category>
            <category><![CDATA[mechanical-engineering]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Fri, 10 Sep 2021 14:15:53 GMT</pubDate>
            <atom:updated>2021-09-10T14:15:53.667Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Research and documentation by Praneet Dighe and team- ISA, Manipal student section.</blockquote><p><strong>General overview:</strong></p><p>Robots are used in places where it out-performs humans and at times where the risk of losing lives is considerable. Bomb detection and disposal requires a high risk of accident and skill for its operation to succeed. Thus, we have been using robots for the past few decades to scout areas for military operations safely, test new explosives, dispose off broken/old explosives. These are not technically Robots but are some machines put together to form a system to do some specific tasks. These tasks include movement of the machine, detecting a bomb, safe disposal of bomb and retreating. Various engineering fields are required to facilitate the above-termed operations, mainly wireless communication technology and micro-controller system connected with a motor arm and the robot’s body for movement.</p><p>Here is an advanced robot used by the military for terrestrial area scouting purposes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/602/1*woNM9Cdm8Ec2CNqKZ3GC9g.png" /></figure><p><strong>The Need for a Bomb Disposal Robot:</strong></p><p>The development of synthetic chemicals has increased the production of powerful explosives. Most of the time, the material with a doubtful origin is manipulated by specially trained bomb squads. However, complicated and even remotely controlled triggering mechanisms challenge a specialist’s talents since the explosive is prepared for tremendous damage. Today’s squads wear safety suits in case of an explosion, which minimizes an explosion’s pressure and prevents injury to a certain level. On the other hand, they limit movement capabilities and increase stress. Despite advanced safety precautions, handling of explosive ordnances brings high risks for the operator’s life. The primary reason and justification of the need for a bomb disposal robot is simply the fact that it saves human life in an explosion. The robot can be operated meters away from the danger zone, providing a totally safe place for explosive specialists. In addition, these robots can be used not only to disarm the explosive ordnance but also to collect information about the threatening material and the area before human interference.</p><p><strong>Application Areas:</strong></p><p>• Handling and manipulation of explosives,</p><p>• Handling of biological, biochemical and other hazardous materials,</p><p>• Manipulation of suspected packages,</p><p>• Neutralizing and disarming unexploded ordnances,</p><p>• Determining and disabling landmines,</p><p>• And collecting visual information in an unsafe environment.</p><p><strong>Movement of the robot:</strong></p><p>Moving machines are not new in today’s world but for starters, it’s basically a machine that includes x motorized wheels, which are remotely controlled and made in a way that supports other sub-systems of machines on it. Various sensors like a Camera and eco-location systems are placed to help better control and navigate around the area. Micro-controller remote is used to control the acceleration and direction, it sends and receives signals through air by infrared signals, Radio frequency and other signals can be used but due to congested environment during operation IR is preferred here.</p><p>From the above basic hardware, it can carry some load and navigate around and hence next step of operation is able to proceed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/323/1*dk7dt1P5L4LI-5sqIhCkNA.png" /></figure><p>Above is an Arduino microcontroller which helps us to control the motors and various other sub-systems of robot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/225/1*_08H6R6Vx58H_c4f0peGEA.png" /></figure><p>This is a 12v brushless dc motor which is efficient and quiet and it is able to produce torque in range of 500–5000 mNm.</p><p><strong>Bomb Detection and Disposal:</strong></p><p>A simple but highly effective system of metal detectors is used to detect suspected area, sometimes various specific sensors like x-ray, infrared sensors are used as a facilitatory system for achieving higher accuracy of detection.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/324/1*oGcVkSzV05i_IxIciElEEg.png" /></figure><p>A miniature metal detection stick is also used in conjunction with other sensors.</p><p>After detection disposal process begins, the oldest method is to spray water using a mounted pressurized water jet over bomb circuit to short it and render it ineffective. It requires a basic pressurized pumping-nozzle spraying system but it increases weight and reduces range of robot.</p><p>We can then retrieve the explosive by physically going to the place or use a mechanical robotic arm that could be mounted on the machine.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/225/1*jROQHJKuhEi3sIV9LI9peA.png" /></figure><p><strong>Sub-system technologies:</strong></p><p>Most of robots are used in wireless and out of view conditions which require various level of wireless and electronic technological hardware and tuning. Basic hardware for enabling wireless function include EMR radiating and catching antennas and its controllers that can be tuned to a remote. This can also add a GPS functionality which makes it much easier to navigate remotely.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/292/1*ZLM2jAKeEbXC-7v5Kmvsdw.png" /></figure><p>This antenna is able to cover about 2 km range with using about 50–75 mW of power and operates at about 400 mHz. This is highly efficient and effective for its range and power consumption.</p><p>Since the operation is wireless, direct power cannot be supplied and hence a High-capacity battery pack to power all sub-systems throughout the operation. It should be chosen with balance of capacity and weight depending on range and time of duration of the operation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/602/1*ZMsq2_N62PXAdm1zJI4z9Q.png" /></figure><p>Batteries of various capacities are used depending on duration of operation, generally a battery with 12v output is necessarily required as minimum since the motors are 12v dc at minimum.</p><p>Areas can be wildly challenging to travel; navigating is essential part of time-sensitive operations. An Electric motor system and transmission system is placed for handling and a static or moving camera is also used for feedback to the operators.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/254/1*o6sWETvoFWbgqQoWsERP1Q.png" /></figure><p>A continuous track system is also generally used for uneven surfaced areas.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/306/1*up1rZ0vEv1jHsb5c_w1AkA.png" /></figure><p><strong>Conclusion:</strong></p><p>The system working coherently with operator while having versatile movement, sensory data and disposal mechanism make it perfectly fit for bomb detection and disposal. In addition, various sensor system for each sub-system can be chosen according to budget and duration of operation makes it much economical and reliable. These are without a doubt a good replacement for human scouts and hence we have seen rise in implementation of this technology by various militaries around the globe.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1c5035507b54" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Fractal Robots]]></title>
            <link>https://isa-manipal.medium.com/fractal-robots-b1740466c428?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/b1740466c428</guid>
            <category><![CDATA[mechanical-engineering]]></category>
            <category><![CDATA[robots]]></category>
            <category><![CDATA[operating-systems]]></category>
            <category><![CDATA[electronics]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Mon, 16 Aug 2021 07:20:08 GMT</pubDate>
            <atom:updated>2021-08-16T07:20:08.360Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>Introduction:</strong></h3><p>A fractal is anything which has a substantial measure of exact self-resemblance that is any part of its body will be alike to the whole object.</p><p>Fractal robots are made of cubic bricks having electronics and mechanical components in them. They can reconfigure themselves and can be controlled by a computer. These machines can change their shape from one object to another. This technology is slowly picking up as it has the potential to work like humans in every field like research, medicine, construction, etc. Fractal robot technology is also known as digital matter control and implementation. Some Fractal Robots also have self-repairing capabilities, thus, minimizing human interference. [1]</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/132/1*EGyjoLG1NLlYjFgw2Y97zg.png" /></figure><p><strong>Construction:</strong></p><p>One of the main goals of the Fractal Robot is to make the design as simple as possible with minimal moving parts so that it can be produced on a mass scale at a really low cost. Generally, the body is made up of plastics and metals. However, ceramics and even clay is used at times to make it more environmentally friendly. The robotic cubes are assembled from face plates. These plates are then bolted to a cubic frame.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/205/1*pkPKujeY-XSxtNcOaWo9RQ.png" /></figure><p>Hence, these cubes are generally hollow. The plates have the mechanisms in them. Each surface of the cube has an electrical contact pad to communicate as well as transmit power to the other robotic cubes. The mechanisms are built into the plates.</p><p>There are slots placed at an angle of 45 degrees on the edges of the cube. Petals are also placed in each face at the same angle and have serrated edges. They are engaged by either a gear wheel or a large screw thread driven by a single motor. Petals can also be driven as a pair by a single motor with the help of a flexible metal strip. Hence, the cubes can slide along.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/255/1*My9EtJAm3I1av5rjKmYlfw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/303/1*jKEb1VPbJhbq3REvJp4HnQ.png" /></figure><p><strong>Computer Control:</strong></p><p>Generally, each robotic cube comes with a microcontroller inside to perform basic operations like control of the internal mechanism and communication. Due to its design, the only movements the robotic cube must do is move left, right, up, down, etc. with respect to each other compared to other types of robots that need complex mechanisms and mathematical modelling. Fractal Robots run on Fractal O.S.</p><p>The Fractal architecture dominates the core functions of the Fractal O.S. All the data that is available to the Fractal O.S. is stored in fractal data structures that permit compatibility and conversions.</p><p><strong>Fractal OS:</strong></p><p>This OS converts the written code into machine commands. The Fractal OS has a number of features like transparent data communication, data compression and awareness of built-in self-repair. Generally, it has RAM, ROM and some counters as a part of its internal design. [2]</p><p><strong>Fractal Bus:</strong></p><p>This permits the software and the hardware to merge seamlessly into one unified data structure.</p><p><strong>Movement Methods:</strong></p><p>There are many designing methods for cubes and are of various sizes, although the movement methods are always similar. The cube obeys instructions to move right, left, forward, backward, up, and down.</p><p>If the cube cannot perform a function, then it turns back. If it cannot perform even that, then the software in the cube starts self-repair algorithms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*8cY31jB6N_JMBdenPfuvdA.png" /></figure><p>Pick and place: instructions are issued for a compilation of cubes telling every cube where to exit. A command of “cube 144 move left by 5 positions” results in only one cube moving in the entire machine. Entire collection of movements needed to perform operations are worked out and stored exactly like conventional robots’ store movement paths.</p><p>N- streamers: A cube is pushed from the outside and an additional cube is shifted into the empty place. The shifted cube is connected to the end of the rising rod &amp; pushed out once more to the rising rod. Can be used for bridge construction.</p><p>L-streamers: L-form cubes are denoted with numbers 4, 5 and 6 and these numbers are connected to a rod denoted with numbers 1, 2 and 3. A new cube ‘7’ is attached so that the rod develops with one cube. The 6 and 7 cubes are shifted to the positions of 5, 6, and 7 to form an L-shape. [3]</p><p><strong>Self-Repair:</strong></p><p>For this system to work, the cube must be partially dismantled and then re-assembled at a custom robot assembly station. The cubic robot is normally built from six plates that have been bolted together.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/266/1*4vxN3YMbU5b-bVOLsU03PA.png" /></figure><p>When large numbers of cubes are involved, these plates (mechanisms) can be stacked onto a conveyor belt system and assembled into the whole unit by robotic assembly station to save space and storage. (By reversing the process, fractal robots can be dismantled and stored away until needed.) If any robotic cubes are damaged, they can be brought back to the assembly station by other robotic cubes, dismantled into component plates, tested and then re-assembled with plates that are fully operational. Potentially all kinds of things can go wrong and whole cubes may have to be discarded in the worst case. But based on probabilities, not all plates are likely to be damaged, and hence the resilience of this system is much improved over self-repair by cube level replacement. [4]</p><p><strong>Applications:</strong></p><p>• Bridge Building</p><p>• Fire Fighting</p><p>• Defense Technology</p><p>• Medical Applications</p><p><strong>Limitations:</strong></p><p>• Technology is still in infancy</p><p>• Current cost is very high ($1000 per cube for the 1st generation of cubes, after which it will reduce to $100 or so).</p><p>• Needs very precise and flexible controlling software [5]</p><p><strong>Bibliography:</strong></p><p>[1], [4]: <a href="http://www.ijritcc.com/">www.ijritcc.com</a></p><p>[2], [5]: <a href="http://www.elprocus.com/">www.elprocus.com</a></p><p>[3]: <a href="http://www.slideshare.net/">www.slideshare.net</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b1740466c428" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Neuromorphic Computation]]></title>
            <link>https://isa-manipal.medium.com/neuromorphic-computation-f198e0d42ebb?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/f198e0d42ebb</guid>
            <category><![CDATA[neural-communication]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[neuralink]]></category>
            <category><![CDATA[science]]></category>
            <category><![CDATA[computational-thinking]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Thu, 29 Jul 2021 13:35:57 GMT</pubDate>
            <atom:updated>2021-07-29T13:35:57.249Z</atom:updated>
            <content:encoded><![CDATA[<h3>INTRODUCTION</h3><p>There has been a tremendous growth in the AI field with the current CPUs and GPUs and the use of supercomputers. However, the present technology lacks machine reasoning, transfer learning, physical dimensions and many more. As a result, we need some serious and major changes in the way we structure a computer. One of those inspiring ways is by making physical computers a little more like a human brain. Neuromorphic chips promise to overcome this challenge with ground breaking research in the field of memristor and artificial synapse.</p><p>It involves designing and engineering computer chips that use the same physics and computation used by our nervous system which is different from an artificial neural network that mimics the logic of how a human brain thinks.</p><p>Neuromorphic computing simulates analog behaviour of human brain. Traditional computers ‘think’ in binary. Everything is either a ‘1’ or ‘0’, a ‘yes’ or a ‘no’. Hence the code we use and the question we ask must be structured rigidly. On the other hand, a neuromorphic computer works a little flexibly.</p><p>To do this we mimic the current flow that happens between neurons via synapses in a hardware. Exploiting the concept of synaptic transmission and flow of information within a network of neurons which are heavily dependent on ionic currents, we try to incorporate multiple states in the neuromorphic chips rather than just a simple “yes” and “no”. Mimicking this ability to transmit a gradient of understanding from neuron to neuron and have them all working together simultaneously results in neuromorphic chips that are more energy efficient especially for complicated tasks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/621/1*6OTcWc-LPPIA3lz5v8YdfA.png" /></figure><p><strong>Artificial synapse:</strong></p><p>The prime aim of a machine learning algorithm is to mimic the ability of the brain to make changes in the form of interpretation of the data based on the need of the surrounding. The brain both transfer and processes electric signal the received from a sensory organ. Inspired by this artificial synapse transfer, store and process the data at the same time without having need to have a separate space for data storage.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/797/1*zb87fjrujvMwZGdeb3Bl4A.png" /></figure><p><strong>Memristor/Memory resistor:</strong></p><p>It consists of the storage layer which is inserted between the electrodes. When a external electrical stimuli is applied to across the layer, it undergoes dynamic reconfiguration within the storage layer resulting in resistance modulation , referred to as memory effect. It has the ability to retain the changed resistance state even after electrical inputs are removed. Thus, it is used for analog switching which resembles biological synapse. The synaptic weight can be increased or decreased depending upon applied potential. In addition to analog switching , memristors also have other desirable properties such as long endurance and retention, nanosecond switching power and low power consumption. Owing to these characteristics memristors has emerged as a promising option for artificial synapse.</p><p>Desirable properties of Memristor synapse:</p><p>Linearity in weight update is one important factor affecting the performance of a memristor It indicates the linear relationship between synaptic weight change and programming pulse. It affects the accuracy of the system as it is associated with the mapping of weights in the algorithm for conductance in memristor synapse. Though most of the memristor shows non linear weight update, where the conductance change gradually saturates.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/787/1*EykA_y-LmQTyxPcJV7RXFg.png" /></figure><p>The resolution capability of storage is influenced by multilevel states and dynamic ranges because numerous conductance states can distinguishably store individual pixels of input patterns. However, in a large scale the device to device variation or the cycle to cycle variation can degrade neuromorphic computing. However, the ability of the neuromorphic architecture to be immune to the variation to some extend is of immense use. The memristor synapse should be able to retain the trained weight after every update of the training phase. Hence, larger the endurance and the retention time more is the achievement of neuromorphic computing. Supervised learning-based networks are less vulnerable to cycle-to-cycle and device-to-device variations. This is because memristor synapses are updated according to calculated errors under known target values. By contrast, the networks based on unsupervised learning are directly affected by the variation owing to unknown target values. Therefore, memristor synapses need to be designed or selected depending on individual neuromorphic networks. Neuromorphic systems based on crossbar array of memristor synapse: In this approach, Al2O3/TiO2−x memristor was used to fabricate a 12 × 12 crossbar array to implement a single-layer network. The 10 input neurons and 3 output neurons are fully linked by 10 × 3 = 30 synaptic weights (Wi,j). Input voltages (Vi = 1…9) assigned from pixels of the 3 × 3 input images were applied to each input neuron. V10 is a bias voltage to control the degree of activation of the output neurons. After being applied into the network, the input voltages were individually weighted depending on each synaptic weight.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/1*-pkHMbPm-7HNqpHLav9K6w.png" /></figure><p>The output neurons received each weighted voltage through linked weights and then integrated the weighted voltages (ΣWi,jVj), where j and i represent the input (j = 1–9) and output (i = 1–3) neurons respectively. The output neurons converted each integrated voltage into output (fi) ranging from −1 to 1 according to the nonlinear activation function: fi = tanh(βIi), where β adjusts the nonlinearity of the activation function and Ii = ΣWi,jVj.</p><p>The synaptic weights were represented by a pair of adjacent memristors (Wi,j = Gi,j+ − Gi,j−) for the effectiveness of weight update. The number of selected memristor synapses in 12 × 12 array were 30 × 2 = 60, due to a pair of memristors . When the network was under the training process, memristor synapses between input and output neurons were updated based on the Manhattan update rule, which is classified as supervised learning: ΔWi,j = ηsgnΣ[(ti(n) − fi(n)) × df/dI × Vj(n)], where η is the learning rate, ti(n) is the target value, fi(n) is the output value, and n is the nth input image. After the training process was complete, the memristor synapses retained their final conductance, and the test process was performed without weight update.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/802/1*kLi1JjwELDL73YAJ0UlWmA.png" /></figure><p>Other notable advances have been done in this field such as:</p><p>A new design has been proposed by a team in the Massachusetts Institute of Technology, which uses single crystalline silicon and silicon-germanium layer on top of one another. As we apply an electric field to this new device provides a well-controlled flow of ions.</p><p>Korean researchers proposed the use of titanium oxide that is even more durable whereas in Colorado they proposed the use of magnets to even precisely the way of computer neuron communication.</p><p>The University of Manchester took a different approach, their system is called SpiNNaker which stands for Spiking neural network architecture.</p><p>They used traditional digital parts connecting and communicating with each other in an innovative way. They used SpiNNaker to stimulate the behavior of the human cortex.</p><p>SpiNNaker is a processor platform optimized for the simulation of neural networks. A large number of ARM cores is integrated in a system architecture optimized for communication and memory access.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*WNdU2AcV55nXShFhqrTwQA.png" /></figure><p>This field has been further studied with the hope that a computer that behave like a brain will give us enough computing power to simulate something as complicated as the task brain does within a fraction of seconds.</p><p><strong><em>Written by Shambhavi Sinha, AI Subsystem at ISA Manipal.</em></strong></p><h3>References:</h3><p><a href="https://www.intechopen.com/chapters/66439">https://www.intechopen.com/chapters/66439</a></p><p><a href="https://iopscience.iop.org/article/10.1088/1757-899X/912/6/062029">https://iopscience.iop.org/article/10.1088/1757-899X/912/6/062029</a></p><p><a href="https://www.researchgate.net/publication/304414471_Development_of_a_neuromorphic_computing_system">https://www.researchgate.net/publication/304414471_Development_of_a_neuromorphic_computing_system</a></p><p><a href="https://arxiv.org/pdf/1705.06963.pdf">https://arxiv.org/pdf/1705.06963.pdf</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f198e0d42ebb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Synchronous Reluctance Motors (SynRM)]]></title>
            <link>https://isa-manipal.medium.com/synchronous-reluctance-motors-synrm-891ea78e02b2?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/891ea78e02b2</guid>
            <category><![CDATA[mechanical]]></category>
            <category><![CDATA[electric-field]]></category>
            <category><![CDATA[magnetism]]></category>
            <category><![CDATA[motor]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Thu, 22 Jul 2021 06:25:56 GMT</pubDate>
            <atom:updated>2021-07-22T06:25:56.072Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>General Overview:</strong></p><p><strong>A </strong>Synchronous Reluctance Motor, commonly known as a SynRM motor, is a type of AC motor that has no windings or magnets on the rotor. In recent years, it has been seen as a viable alternative to induction motors due to its rigid construction, lost cost, and maintenance. Due to the magnetic locking between the magnetic fields of the stator and rotor, the motor runs at synchronous speed. In terms of performance, it has a fast dynamic response, higher efficiency, power density, and torque.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z0WLKqxAPMUR7VLV8zQzIg.jpeg" /></figure><p>Synchronous reluctance motor combines the performance of the permanent magnet motor along with the simplicity and service-friendliness of an induction motor. [1]</p><p><strong>About SynRM:</strong></p><p><em>Construction:</em></p><p>Whether it runs on AC or DC, any motor comprises two main components: a stator (stationary component) and a rotor (rotating component). The stator of a synchronous reluctance motor has distributed and concentrated winding. The windings usually are three-phase symmetrical in arrangement and can produce a sinusoidal wave in the air gap between the stator and the rotor. The stator has projecting electromagnetic poles on it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/1*XY2BM6RqBvOcrPhMRC_ixQ.jpeg" /><figcaption>Stator</figcaption></figure><p>The rotor has no windings or magnets. Instead, it has electric steel plates/iron laminations stacked together to make a rotor package. Laminations reduce hysteresis losses (caused by magnetization saturation). A nonmagnetic material separates each lamination. Sheets of ferromagnetic and nonmagnetic steel are bonded. The bonded sheets are first cut into rectangular blocks, which are then machined into the rotor. The air gap between the rotor and the stator is uneven. This is done because as the air gap increases, the reluctance decreases. Generally, SynRM comes with three types of rotors. They are salient poles, axially laminated rotor, transversally laminated rotor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/260/1*i4bA-Wk6XXViST8uMBYigw.jpeg" /><figcaption>transversally laminated rotor.</figcaption></figure><ol><li>Rotor with salient poles: In this arrangement, the quadrature air gap is much larger than the direct air gap, which is useful for high-speed application.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/352/1*6dELAYVhMuEV5EzjTu2Dxg.jpeg" /></figure><p>2. Axially laminated rotor: The torque ripple and iron losses are more in axially laminated rotor than transversally laminated.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/242/1*eKhme6MT44ozzgCZNoSoVw.jpeg" /></figure><p>3. Transversally laminated rotor: The lamination required for the shaft weakens the rotor. It is used for low-speed design. [2]</p><p><em>Working:</em></p><p>In a SynRM motor, a magnetic field is created by a sinusoidally distributed stator winding which rotates at synchronous speed and is sinusoidal. Once alternating current is passed through the stator windings, it creates a rotating magnetic field in the air gap of the electric motor. As the rotor tends to establish its most magnetically conductive axis (commonly known as d — axis) with an applied field in order to minimize the reluctance in the magnetic circuit, torque is generated. This torque generated is called the reluctance torque.<strong> </strong>Reluctance torque is the torque generated because the motor is moving to a position where the reluctance seen by the armature flux is declining. The more significant the difference between the direct inductance and flux linkage per armature ampere, the greater the torque generated.</p><p>If the reluctance torque is sufficient to start the motor, the rotor will pull into step with the revolving field and continue to run at the speed of the revolving field that is of the stator.</p><p>When a stator pole is energized, the rotor torque is in the direction that will reduce reluctance. Thus, the nearest rotor pole is pulled from the unaligned position into alignment with the stator field (a position of less reluctance). In order to sustain rotation, the stator field must rotate in advance of the rotor poles, thus constantly “pulling” the rotor along. Some motor variants will run on 3-phase AC power. Switched reluctance type are primarily used in modern-day designs because significant control advantages for motor starting, speed control, and smooth operation are provided by electronic commutation. [3]</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/814/1*Q09iZ20XaWZDpSFGPSGI0g.png" /></figure><p><strong>Advantages:</strong></p><ul><li>Unlike synchronous motors, which require starting methods, SynRM doesn’t require any starting method due to the reluctance principle. It is self-starting.</li><li>It is robust and has a less complicated structure; hence low maintenance is required.</li><li>It does not heat up during operation since there are no currents in the rotor, thus increasing the service life of the motor.</li><li>The absence of windings and magnets on the rotor results in moment of inertia of the rotor being lower, allowing the electric motor to accelerate faster and save energy.[4]</li></ul><p><strong>Applications:</strong></p><ul><li>Tesla’s Model 3 has this type of motor in use!</li><li>They are applicable in low-power applications like fiber spinning mills due to their low cost, inherent simplicity, and construction is robust, etc.</li><li>They are used in applications where constant speed is required, like timing devices, phonographs, control devices, recording instruments, conveyor belts, rice mills, etc.</li><li>It is used like proportioning devices within conveyors or pumps.</li><li>They are also used in the process of film material or continuous sheet folding and wrapping machines. [5]</li></ul><p>Bibliography:</p><p>[1] <a href="http://www.lesics.com/">www.lesics.com</a></p><p>[2] <a href="https://en.engineeringsolutions.ru/">https://en.engineeringsolutions.ru/</a></p><p>[3] <a href="http://www.watelectrical.com/">www.watelectrical.com</a></p><p>[4] <a href="https://en.engineering-solutions.ru/">https://en.engineering-solutions.ru/</a></p><p>[5] <a href="https://www.inverterdrivesystems.com/">https://www.inverterdrivesystems.com/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=891ea78e02b2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Canadarm]]></title>
            <link>https://isa-manipal.medium.com/canadarm-ba02cc19c578?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/ba02cc19c578</guid>
            <category><![CDATA[nasa]]></category>
            <category><![CDATA[internationalspacestation]]></category>
            <category><![CDATA[space]]></category>
            <category><![CDATA[deepsky]]></category>
            <category><![CDATA[satellite-technology]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Thu, 08 Jul 2021 11:55:38 GMT</pubDate>
            <atom:updated>2021-07-08T12:39:13.854Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>This article is presented by ISA Manipal in collaboration with Parikshit student satellite team, Manipal Institute of Technology, Manipal.</blockquote><h3>What is Canadarm?</h3><p>The Canadarm also called the Shuttle Remote Manipulator System (SRMS) is basically a robotic manipulator which is 15 meters in length consisting of various components. It weighs 410 kg. It had 6 degrees of freedom (Canadarm (2) has 7). It has a maximum payload of 266,000 kilograms in space and can operate at speeds of 60 cm/s. It can place its payload within 5 cm of the required target. Canadarm played a significant role in helping set up the International Space Station. After 30 years of active service, the Canadarm was retired and replaced by Canadarm 2, which is currently attached to the ISS and helps in assisting the capsules that deliver cargo. It is also used to help astronauts and assists them during spacewalks repairs and upgrades on the outside of the ISS.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/624/1*1LvcqucTu4AeaFDpEkiOkA.png" /></figure><h3><strong>Materials used:</strong></h3><p>The manipulator is made up of titanium, stainless steel and ultra-high modulus graphite-epoxy. These materials meet the strength requirements. In order to protect the arm from harsh thermal conditions, a multi-layer insulation system consisting of alternate layers of goldized Kapton, Dacron scrim cloth and a fiberglass outer covering. For the electronics to withstand extremelycold temperatures, thermostatically triggered electric heaters are provided within the body of the manipulator.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*LCzJ2pGFVonb5TKNjX4tOg.gif" /></figure><h3>Working Mechanism:</h3><p>The Canadarm has 6 joints. Two in the shoulder, one in the elbow and three in the wrist of the robotic arm. They are all made up of an element known as Joint One degree of freedom (JOD). JODs are gear boxes that move and work the same as a normal human arm. The gear boxes are motor driven. The joint present in the elbow helps the arm to pitch while the joints present in the wrist of the robotic arm allows it to pitch, yaw and roll. The speed of the motor can be controlled, so as the maneuverability of the arm. The motors also have breaks and are capable of moving in different directions. Each joint is driven by a servo mechanism. The output of this mechanism is provided by a brushless DC motor and finally a high resolution gearbox is used to transmit the output. Normally brakes are provided in the joints and are applied when the joints are un-commanded.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/624/1*LD-6Y48cBzhz5CtuNzx0Tg.png" /></figure><h3><strong>End Effector:</strong></h3><p>As the manipulator would be operating in space under zero gravity conditions, the end effector cannot have traditional jaw-like grippers to capture stationary or free flying payloads. The end effector of the Canadarm provides a large capture envelope of 20.3 cm in diameter and 10 cm in depth. The structure and mechanism of the end effector is capable of providing a soft docking and rigidizing. The end effector accomplishes this by a 2-stage mechanism. It closes 3 cables in the form of a snare around a knobbed pin which is bolted onto a payload. It then draws it to the device until close contact is established. This then applies a load of approximately 499 kg onto the knobbed pin. Hence, allowing the astronauts to maneuver the payload.</p><h3>How is the Canadarm controlled?</h3><p>The main controls of the Canadarm are in the cupola module of the ISS. Cupola is the module that lets the astronauts have a 360-degree view of the outside of the ISS, including the view of the Canadarm.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/317/1*7v9S2arI9f6JHUnUxC5OUA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/292/1*A7goCjlSyEmtaykFH9ULEA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/472/1*nIIdWjhh9jxA6xxP2rME9A.png" /></figure><p>Inside the cupola, there are:</p><p>● 4 displays- which show various data that is crucial to controlling the arm, including the view from the on-board cameras.</p><p>● One joystick- That is used to manipulate the rotational properties.</p><p>● One additional control to manipulate the translational position of the arm.</p><p>● There is also an additional panel that can be used to toggle different cameras and modes for the Canadarm.</p><p>The Canadarm can also be controlled remotely from the NASA and Canadian space agency mission centers as and when needed.</p><p>As an important additional benefit, this mode of operation allows astronauts to work on the shuttle without having to wear a spacesuit. The arm is also designed as a support platform that allows astronauts to repair satellites while wearing extravehicular activity (EVA) spacesuits.</p><p>To help increase the redundancy, every joint has 2 motors, along with two electronic circuitry at each point. Being in space, outside the gravity’s influence, the entire ISS is prone to radiation from various sources. Although this is rare, this radiation activity has the capacity of corrupting the flight system hardware. To counter this, special components that are ‘red hardened’ are used. Multiple components are also used to prevent data corruption.</p><p>A few advancements of the Canadarm 2 is that they have 3 extra sensors to facilitate a safer working environment in space. They are:</p><p>a. Force moment sensors that provide a sense of touch</p><p>b. Cameras and other sensors that enable automatic vision feature for capturing modules.</p><p>c. There are multiple sensors that are used to implement a collision avoidance system.</p><h3>Arm Control System:</h3><p>The movement can be controlled by the controls linked to the general-purpose computer (GPC). The hand controllers used by the astronauts are used to tell the computer what the astronauts would like the arm to do. The software then uses the predefined functions and kinematic equations to figure out which joints to move, by how much and how fast to move them. Every change is updated to the GPC every 80 milliseconds. The computer then recalculates the above factors (joints to move, the speed and how much) and sends the updated commands based on the astronaut’s requests.</p><p>The arm health is also monitored every 80 milliseconds, and if a failure occurs in the algorithm, the entire arm ceases its motion to enable rectification. If the task at hand is important, there is also an option at the flight control deck, to control each joint individual.</p><p>If everything fails, there is a final safety feature is a jettison capability to ensure that the payload bay doors are always closed</p><h3>Arm Instrumentation:</h3><p>The current instrumentation present on the arm includes</p><p>● A joint angle optical encoder that gives an accurate measure of joint position. They are mounted on the gearbox output shaft of each joint. The encoders used are 16 bit encoders to get precise values.</p><p>● A joint tachometer which provides joint rate data (sends data in the form of 12 bits) and strain gauges to measure the bending and torsion. They are located on the motor output shaft. And the strain gauges are mounted near the shoulder pitch and wrist pitch joints.</p><p>● There are also 6 closed circuit television cameras (CCTV) to primarily facilitate payload handling.</p><p>The CCTVs provide 9x zoom, a pan capability of 170 degrees on both sides, and a tilt capability of 170 degrees on both sides as well</p><p>The tachometer and the encoders record data at a rate of 12.5 Hz to upload it to the control deck. The strain gauge, on the other hand, has a transmission frequency of 25Hz</p><p>The arm is also said to have a Force-Torque Sensor sandwiched between the wrist roll joint and the end effector, and is useful to provide the torque and force data while the arm is in use. This data can then be utilized for future analysis</p><p>· The altitude control system of the spas contains 3 linear accelerometers which are Sundstrand QA 1200-AA 08 style accelerometers and have an operating range of 0.1g to 0.0001g</p><h3>The 6 Degrees of Freedom:</h3><p>The inflight ‘<em>6 Degrees of Freedom</em>’ control system of the Canadarm had two parts in it:</p><p>● <strong>Control by Onboard computers</strong></p><p>● <strong>6 Degree of Freedom Control by the two in hand controllers.</strong></p><h3>1. Control by Onboard computers:</h3><p>The controller onboard the computer is operated by a technology named “<em>Fly by Wire”</em> (<strong>FBW</strong>).</p><p>An FBW control system is one that places a computer between the operator and the actuators in the arm’s joints. Fly-by-Wire (FBW) is the common term for flight control systems that employ computers to process the pilot or autopilot’s flight control inputs and send corresponding electrical signals to the flight control surface actuators.</p><p>The advantage of this technology is that such a control system is relatively light in weight, can be re-programmed to meet different requirements and now, not as costly as other alternatives. As previously noted, there was a backup mode of operation that allowed the Canadarms’ joints to be operated individually. As a matter of interest, these FBW control systems are now common in modern aircraft design and are also being introduced in automobiles for acceleration, braking and steering.</p><h3>2. The working principle of a “Fly by wire” control system:</h3><p>The FBW control system is a method of controlling an actuator without any human involvement and decisions following the input of a control step. Eg: Moving the joystick to the right for a right turn. After this input the onboard system figures out the kind of input to feed the actuators with to get the desired output. <br> The system guides its own self by taking inputs in the from of control variables from the onboard sensors in a feedback loop to achieve real time correction and inputs for the next cycle of the control flow, A similar controller is employed onboard aircrafts which have gyroscopes for sensing the angular velocity and the onboard algorithm outputs the torque the actuators have to give in order to correct the orientation.</p><h3>3. 6 Degree of Freedom Control by in-hand controllers:</h3><p>The two in-hand controllers were hand-operated by the astronauts from the view they had through the large windows in the space shuttle and the cameras on the elbow and wrist of the arm. These motions were pre-programmed on the arm’s onboard computer already.</p><p>The canadarm was the first space robotics of any size. An extremely clever breakthrough was the control by a joystick. The tip was controlled by the joystick, and all the joints were coordinated to make the same happen. The six degrees of freedom, similar to a human arm are as follows:</p><p>● Two joints in the shoulder</p><p>● One joint in the elbow</p><p>● Three joints in the wrist</p><p>● Elbow rotation, limited to 160 degrees</p><p>The cameras placed on the elbow and wrist assisted the astronauts on the space shuttle to control the robotic arm</p><h3>Thermal Protection System:</h3><p>Since the arm is fixedly installed in the cargo compartment of the space shuttle, methods must be developed to control the sub-zero temperature on the shadow side of the arm and the extreme heat on the sunlit side. The solution is to cover the entire arm with a multi-layer thermal insulation blanket to provide passive thermal control. This material is composed of alternating layers of products manufactured by DuPont. It has unique electrical, thermal and chemical properties and can reflect heat and cold. In addition, the electric heater is connected to key mechanical and electronic components and turns on and off in sequence to maintain a stable operating temperature.</p><h3>How does docking work?</h3><p>The grappling fixture that is employed is known as the Latching End Effector (LEE). This consists of three high tension wires that are controlled by 3 separate disks in such a way that they form a triangle once they are in the ‘docking’ mode. There are multiple sensors and motors involved here that make sure that the docking part is done as smoothly as possible.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/289/1*9m3KEFzYuctGdbFn209bWw.png" /></figure><h3>Safety Strategies:</h3><p>There were multiple fail-safe procedures that were decided on and coded into the software to ensure the safety of the astronauts on board. The most important ones out of them are:</p><p>● Flexible modes for the arm</p><p>● Dynamic fatigue protection</p><p>● Collision avoidance</p><p>● If everything were to fail, there was also the option to control each individual joint of the arm separately to get the arm in a safe position to carry out repairs</p><p>All the parts of the arm are modular to facilitate easy repairs while in space.</p><h3><strong>Canadarm 1 To Canadarm 2:</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/597/1*QDnb3YTRwidUiEeNtHeOIA.png" /></figure><h3>References:</h3><p><a href="https://eic-ici.ca/honours_award/cit12/arm.pdf">https://eic-ici.ca/honours_award/cit12/arm.pdf</a></p><p><a href="https://gravitymovie.fandom.com/wiki/Canadarm">https://gravitymovie.fandom.com/wiki/Canadarm</a></p><p><a href="https://www.thecanadianencyclopedia.ca/en/article/canadarm">https://www.thecanadianencyclopedia.ca/en/article/canadarm</a></p><p><a href="https://www.youtube.com/watch?v=K7NvsxcoDKo">https://www.youtube.com/watch?v=K7NvsxcoDKo</a></p><p><a href="https://www.youtube.com/watch?v=cRt8cH1iMp4">https://www.youtube.com/watch?v=cRt8cH1iMp4</a></p><p><a href="https://www.nasa.gov/mission_pages/station/structure/elements/subsystems.html">https://www.nasa.gov/mission_pages/station/structure/elements/subsystems.html</a></p><p><a href="https://www.innovation.ca/story/canadarm-technology-helps-manufacturers-be-more-efficient">https://www.innovation.ca/story/canadarm-technology-helps-manufacturers-be-more-efficient</a></p><p><a href="https://documents.techno-science.ca/documents/CASM-Aircrafthistories-SparMDACanadarms.pdf">https://documents.techno-science.ca/documents/CASM-Aircrafthistories-SparMDACanadarms.pdf</a></p><p><a href="https://ntrs.nasa.gov/api/citations/19900007820/downloads/19900007820.pdf">https://ntrs.nasa.gov/api/citations/19900007820/downloads/19900007820.pdf</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ba02cc19c578" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Magnetic Levitation Train.]]></title>
            <link>https://isa-manipal.medium.com/magnetic-levitation-train-fb98262eebea?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/fb98262eebea</guid>
            <category><![CDATA[magnetism]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[trains]]></category>
            <category><![CDATA[isa]]></category>
            <category><![CDATA[levitation]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Thu, 01 Jul 2021 14:22:48 GMT</pubDate>
            <atom:updated>2021-07-01T14:22:48.604Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>General overview:</strong></p><p>Magnetic levitation train is also known as Maglev train is a floating train used for land transportation that’s supported by either electromagnetic attraction or repulsion.</p><h3>Two sorts of maglevs are in commission Electromagnetic suspension (EMS) and electrodynamic suspension (EDS). [1]</h3><p><strong>How does it work:</strong></p><p>These trains float over guideways using the principles of magnets. There isn’t any rail friction, meaning these trains can hit speeds of many miles per hour.</p><p>There are three components in a Maglev train system:</p><p>1. A large electrical power source</p><p>2. Metal coils lining a guideway or track.</p><p>3. Large guidance magnets attached to the bottom of the train.</p><p>The magnetized coil run along the guideway. [2]</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/290/1*k7wI8vJXWYWoaDVTEsImJw.png" /></figure><p>There are three varieties of levitating systems:</p><p>· Electromagnetic Suspension</p><p>· Electrodynamic Suspension</p><p>· Inductrack System</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*PS34JtHwVsL00Nx8b0LWrQ.png" /></figure><p>1.Electromagnetic suspension (EMS): It uses the attractive force between magnets present on the train’s sides and underside and on the guideway to levitate the train. In this system, the bottom of the train surrounds a <a href="https://science.howstuffworks.com/iron4.htm">steel</a> guideway. Electromagnets attached to the train’s undercarriage are directed up toward the guideway, which levitates the train by about 1 inch above the guideway. Additional guidance magnets embedded in the train’s body keep it stable during travel.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/224/1*X0L5rQPmq9d9yk0UyQcbfw.png" /></figure><p>It is simpler to implement than Electrodynamic Suspension and it maintains levitation at zero speed. The drawbacks are that the system is inherently unstable even though it has guidance magnets. This causes the train to not levitate and come grinding to a halt.</p><p>2. <strong>Electrodynamic suspension (EDS):</strong> uses magnets to repel the train from the guideway rather than attract them. The electromagnets used are super-cooled and superconducting. This type of electromagnet can conduct electricity even after the electric current supply is cut off. Cooling the coils to a very low temperature also helps to save essential energy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/215/1*N0wdUxMnoGfuh--lVIIp-w.png" /></figure><p>This type of levitation helps maintain stability even at high speeds. Maintaining correct distance between train and guideway is not a major concern. The main drawback is that sufficient speed needs to be built up in order for the train to levitate. Also, this system is complex and expensive.</p><p><strong>3. Inductrack Levitating system:</strong> The Inductrack is a newer type of levitating system that uses permanent room-temperature magnets to produce the magnetic fields instead of powered electromagnets or cooled superconducting magnets. A power source is used to accelerate the train only until it begins to levitate. In case of power failure, the train can slow down gradually and stop on its auxiliary wheels.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/215/1*WdhEITkAv-BmC2cNXdCjpw.png" /></figure><p>The track contains electrically shorted circuits containing insulated wire in an array. When the train moves, a magnetic field repels the magnets which causes it to levitate. The magnets are configured so that the intensity of the magnetic field concentrates above the array rather than below it. This helps to create enough levitating force.</p><p>[3]</p><p>The magnets employed are once again superconducting here.</p><p>Once the train is levitated, power is supplied to the coils within the guideway walls to form a novel system of magnetic fields that pull and push the train along the guideway. The electrical current supplied to the coils within the guideway walls is alternating to change the polarity of the magnetized coils after every half cycle. This variation in polarity causes the magnetic field in front of the train to drag the vehicle forward, while the magnetic field behind the train adds more forward thrust.</p><p><strong><em>Propulsion:</em></strong></p><p>Maglev uses an electrical linear motor for propulsion. The stator is laid flat and therefore the rotor rests above it. Rather than a rotating magnetic flux, the stator generates a field that travels down its length.</p><p>In a linear motor, the stator is commonly called as the ‘primary’ and the rotor as the ‘secondary’. In maglev trains, the secondary is attached to the underside of the train cars, and thus, the primary is in the guideway. So, a field is sent down the guideway and it pulls the train along after it. However, in this configuration, the secondary always lags behind the moving field in the primary. This ends up in speed loss. In a Linear Synchronous Motor (LSM), this lag is removed by attaching permanent magnets to the secondary. Since, the secondary is now producing its own stationary magnetic field, it travels down the primary in sync with the moving field. Hence, due to increased efficiency, LSMs are the preferred choice.</p><p>The aerodynamic shape of the Maglev train allows it to reach speeds of up to 500 km/hr.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/290/1*OtdJiSyINaahL3gYqsLteQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/377/1*EIQ-K3YtwqnqdHwGwEmvlw.png" /></figure><p><strong>Guidance:</strong></p><p>Repulsive magnetic forces help in achieving guidance in high-speed maglev trains. There are two sorts of guidance namely Trans Rapid guidance and MLX guidance used for Maglev trains.</p><p>In the Trans Rapid guidance, two electromagnetic rails are placed on the train which face either side of the guideway.</p><p>In the MLX, guidance is integrated with the levitation system. The levitation rails on either side of the train are connected to each other to ensures that the train stays on course. [4]</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*fyJWNKAilNIW1JFv3XZJwA.png" /></figure><p><strong>Purpose:</strong></p><p>These trains have a high speed. Because the trains rarely touch the track, there’s far less noise and vibration than traditional trains. Due to less friction and mechanical vibrations fewer mechanical breakdowns will tend to occur, meaning that maglev trains are less likely to encounter weather-related delays.</p><p>The Maglev train can reach speeds of about 500 km/hr and some even up to 601 km/hr. These speeds will allow for faster mass and efficient mass transport.[5]</p><p>Bibliography</p><p>[1] <a href="https://www.britannica.com/">https://www.britannica.com/</a></p><p>[2] <a href="https://science.howstuffworks.com/">https://science.howstuffworks.com/</a></p><p>[3] <a href="https://www.britannica.com/">https://www.britannica.com/</a></p><p>[4] <a href="https://sites.tufts.edu/">https://sites.tufts.edu/</a></p><p>[5] <a href="https://www.energy.gov/">https://www.energy.gov/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fb98262eebea" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Neural Networks]]></title>
            <link>https://isa-manipal.medium.com/neural-networks-2c07b53094df?source=rss-e1741a63268c------2</link>
            <guid isPermaLink="false">https://medium.com/p/2c07b53094df</guid>
            <category><![CDATA[cosmology]]></category>
            <category><![CDATA[climate-change]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[automation]]></category>
            <dc:creator><![CDATA[ISA Manipal]]></dc:creator>
            <pubDate>Thu, 27 May 2021 16:55:37 GMT</pubDate>
            <atom:updated>2021-05-27T16:55:37.628Z</atom:updated>
            <content:encoded><![CDATA[<h4>Leveraging Artificial Intelligence to further our own.</h4><p>One of the most intriguing creations in the field of machine learning and at the very centre of deep learning, neural networks are an artificial method of mimicking the human brain’s biological actions of sending neural signals and identifying patterns. The different “neurons” within the network each receive certain information, process it, and then pass it forward from which, at the end, a certain conclusion is obtained. This information can then be considered as a certain form of intelligence. In this paradigm, a computer is usually fed with various labelled inputs in which it recognizes patterns and is then able to accurately correlate new inputs with what their classifications should be.</p><h3>Neural Networks vs Climate Change</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/506/1*Pud1elLLdQNhdVbepPLSSQ.png" /></figure><p>The application of these artificial neural networks can be found in all parts of our daily lives such as language translators, face recognition software or even the customized social media feeds we meet every day; however, its application can be seen on a much larger scale as well. A neural network by the name of Calving Front Machine (CALFIN) has been created which can analyse decade’s worth of satellite imagery to measure the glacier edges and their changing sizes, with the same accuracy as trained scientists. It is dependable, autonomous and can monitor the rate of glacier loss taking place all over the world by processing countless amounts of glacier images, which humans cannot. To prepare the neural network, a training period was conducted where the network was exposed to thousands of glacial images and the weights and thresholds for each of the nodes were tweaked accordingly to achieve the final network which could judge the new inputs correctly.</p><h3>Neural Networks in Cosmology</h3><p>Another application, reported through an article by Carnegie Mellon University, is the use of neural networks and machine learning to better compute cosmological simulations. Cosmological simulations are an essential source of information as they allow scientists to analyse data about dark matter and dark energy in the universe. However, the issue that the researchers faced was that they could either have low-resolution simulations of a large-scale view or a high-resolution simulation of a very small-scale view, which prevented them from being able to analyse all areas and aspects accurately. To counter this obstacle, professors at Carnegie Mellon University have created a machine-learning algorithm that can convert a low-resolution simulation into a super-resolution simulation. The algorithm can generate simulations that can focus clearly on up to 512 times as many more particles, all under a reduced processing time that is only a tenth of what it was originally.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/612/1*_2DbZ-5q6q4Sr_65G7vZSw.png" /><figcaption>A schematic depicting a GAN (Generative Adversial Network).</figcaption></figure><p>The Professors that are heading this project, Yin Li and Yueying Ni, expect the algorithm to get much more efficient as it is continuously trained using various data sets. They used a specific approach known as the generative adversarial network, in which two neural networks are pitted against each other. Given some inputs, both the networks are used to create new data to arrive at a result with more speed and accuracy. In this case, one of the neural networks handles taking low-resolution images as input and converting it into high-resolution images. The second network handles distinguishing between the artificially created image and the image created conventionally by scientists. The method is repeated until the second network is unable to distinguish between the artificial simulation and the real one created by humans, meaning that the first network is creating simulations with just as much accuracy as normal scientists.</p><p>But the blazing efficiency we see today in the creation of such simulations was not easy. The algorithm did not work for over 2 years until finally, the Professors made a breakthrough and the model started generating accurate models. These simulations were focused on dark matter and gravity primarily and hence have omitted “smaller” events such as star-formation and the effects of black holes. They hope to extend the algorithm to study these events as well in the future. Studying such a large entity such as the universe can seem impossible, but with breakthroughs like these, we get closer to understanding the workings of the universe.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/624/1*p6ytxHGXt65uk_2_080RxQ.png" /></figure><h4>How it works</h4><p>The actual working of the network is such that it is made up of a series of smaller nodes that are densely interconnected and may even have various layers. Each of these nodes is connected to a multitude of other nodes from which it receives data or sends data. The nodes all have a certain value called a “weight” and whenever they receive certain information, they process it using the weight and then compare it to the pre-set threshold value. If the processed value is smaller than the threshold, no value gets passed onwards from the node, but if it is greater than the threshold, the node pushes forward its number output, which is a sum of all its weighted inputs, to all its outward connections.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/624/1*fO1miL0hSaWi2QxA5JBzOw.png" /></figure><p>Mathematically, first of all, each input is multiplied with the corresponding weights that have been assigned to that specific node. The weight signifies a particular correlation between two particular nodes and decides how much/little they influence each other, and it is evaluated through the following summation equation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*ToY0ZlgDuUzqRINKaaIldg.png" /></figure><p>Next to obtain the row vectors of inputs and weights, the dot product is calculated:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/508/1*iePB-rjGSxh8bdpJZnUWow.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/110/1*8wtUicjzBQeENHMP0KNSPg.png" /></figure><p>Adding the bias of that node, the function is moved forward to obtain the required output:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/135/1*hlvN5kQlbDh0aaSw6Pt2SA.png" /></figure><p>The last step in this process of forward propagation is passing the value obtained through an activation function, such as the sigmoid function, which makes it undergo a nonlinear transformation. This makes it better suited to handling a variety of different complex scenarios which may not be linear in nature. The sigmoid function for example is:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/215/1*H3aUSjvlnD7cBCItu_2oBA.png" /></figure><p>Next, to understand the accuracy of the network, the difference between the actual value of the input and the value predicted by the node is analyzed using Mean Squared Error formula.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/204/1*8Sn4e-tiGr3hmT24Y35q2g.png" /></figure><p>Once this accuracy is found, it must be understood by how much the algorithm is overshooting or undershooting the predictions and based on that the weights and the bias are changed accordingly to better enhance the working of the algorithm, in a sense allowing it to “learn.”</p><p>A certain Cost function is used to understand the loss in accuracy, where the lower the Cost function, the lower the difference between the actual and predicted values. Therefore, in every iteration, the partial derivative of that cost function is calculated with respect to each of the parameters in the equation as such:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/256/1*jHpm531CuNuFF0r-uudRYw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/515/1*Ig16jp-JpVKqyRS6hvN1gg.png" /></figure><p>This can be simplified to:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/252/1*turRvDXq7iochIENMcZKQg.png" /></figure><p>And the values of the bias and the weights are updated accordingly with every iteration to reach a value as accurate as possible.</p><h3>Conclusion</h3><p>In a sense, neural networks have opened us up to a world of new possibilities in terms of a very accurate and surprisingly fast analytics system with skills that can almost match that of a human. In today’s day and age where information and knowledge are essentially the key to any step forward, neural networks are sure to take up an indispensable role in all fields and industries, prompting continued research and a deeper understanding of their working and implementations.</p><p><em>Written by Ishita Pandey and Yaswanth Biruduraju</em></p><h3>References</h3><ul><li><a href="https://news.mit.edu/2017/explained-neural-networks-deep-learning-0414">https://news.mit.edu/2017/explained-neural-networks-deep-learning-0414</a></li><li><a href="https://www.ibm.com/cloud/learn/neural-networks">https://www.ibm.com/cloud/learn/neural-networks</a></li><li><a href="https://news.climate.columbia.edu/2021/05/05/artificial-neural-network-joins-fight-against-receding-glaciers/">https://news.climate.columbia.edu/2021/05/05/artificial-neural-network-joins-fight-against-receding-glaciers/</a></li><li><a href="https://www.cmu.edu/mcs/news-events/2021/0504_supersims.html">https://www.cmu.edu/mcs/news-events/2021/0504_supersims.html</a></li><li><a href="https://towardsdatascience.com/introduction-to-math-behind-neural-networks-e8b60dbbdeba">https://towardsdatascience.com/introduction-to-math-behind-neural-networks-e8b60dbbdeba</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2c07b53094df" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>