<?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 Varun Mani on Medium]]></title>
        <description><![CDATA[Stories by Varun Mani on Medium]]></description>
        <link>https://medium.com/@varun_mani?source=rss-29643553e9------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*xg4eX8qIMovtXTUo.PNG</url>
            <title>Stories by Varun Mani on Medium</title>
            <link>https://medium.com/@varun_mani?source=rss-29643553e9------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 01:56:04 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@varun_mani/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[Spatial Data: The next frontier of business intelligence]]></title>
            <link>https://medium.com/@varun_mani/spatial-data-the-next-frontier-of-business-intelligence-c77e6cb19504?source=rss-29643553e9------2</link>
            <guid isPermaLink="false">https://medium.com/p/c77e6cb19504</guid>
            <category><![CDATA[business-intelligence]]></category>
            <category><![CDATA[augmented-reality]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[spatialdata]]></category>
            <dc:creator><![CDATA[Varun Mani]]></dc:creator>
            <pubDate>Mon, 09 Jul 2018 21:31:23 GMT</pubDate>
            <atom:updated>2018-07-09T21:31:23.282Z</atom:updated>
            <content:encoded><![CDATA[<p>In a world where digital transformation has become such a hot topic for business leaders, companies and industries that have made the switch quickly have seen disproportionately larger benefits in their top and bottom lines<a href="#_ftn1">[1]</a>. The industries that have lead this charge, and understandably so, have been the ones that are inherently digital or whose operations lend themselves to a smooth digital transformation. Industries like software engineering, financial tech (Fin-Tech), and media and entertainment have assets that are inherently digital and workers that spend most of their time at a desk, in front of a computer. Unfortunately, industries on the other end of the spectrum, like manufacturing and oil &amp; gas have a harder time going through a digital transformation because of the inherent physical nature of the assets and processes involved. Most workers in these industries perform tasks in the physical world as opposed to on computers and these front-line workers amount to over 2 billion, or three-quarters, of the entire global workforce<a href="#_ftn2">[2]</a><a href="#_ftn3">[3]</a>. Unfortunately, the digital tools that their desk-based counterparts have been using for decades have yet to permeate this group.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/514/1*rcFOEXM7NQI-7OjApQ996g.png" /></figure><p>The idea of a digital twin<a href="#_ftn4">[4]</a> helps bridge this gap by creating digital versions of physical assets. Initially coined by DARPA, a digital twin is a dynamic, digital model of a physical thing or system. Abstracting the physical world down to digital representations allows us to leverage cheap and readily available computational power to run simulations and find patterns in our world. Today, businesses can query their Customer Relation Management and Enterprise Resource Planning systems to get in-depth analyses about their customers and sales pipeline. They can rapidly sift through terabytes of usage data to plan out product road maps and can use historical data from millions of sensors to predict when a deployed product needs servicing. In fact, one can simplify digital transformation in businesses down to just four parts: data sources, data lakes, analytics and insights<a href="#_ftn5">[5]</a> as described in a recent HBR article exhibit (reproduced above). Industries that have inherently digital data sources have a much easier time going through a digital transformation while industries with frontline workers are still catching up since their core business-related data sources are just starting to get digitized through sensors and Internet of Things (IoT) platforms.</p><p>This article posits a new type of data, called spatial data. Spatial data enables industries with front-line workers to tap many new data sources and adds a layer of context to their existing data sources by precisely tracking their locations in space. Today, the digital twin concept creates digital representations of <strong>assets</strong> used in a workplace; however, there is still no way to represent the <strong>people</strong>, <strong>places</strong> and <strong>processes</strong> that interact with these assets daily. With the added layer of spatial data, a business’s digital transformation can finally extend beyond physical objects to everything else that front-line businesses care about.</p><p>Spatial data is not a new concept. In fact, the Global Positioning System (GPS) is a great example of large-scale spatial data that has transformed the way many businesses like Uber, FedEx, Amazon and Maersk run. Unfortunately, the precision tracking that GPS provides cannot be used indoors since it relies on communicating with GPS satellites in orbit. Thus, it has never been possible to leverage spatial data for any business processes that occur indoors. Fortunately, the advent of augmented and virtual reality (AR/VR), has finally given us breakthroughs in indoor tracking capabilities that meet and even exceed the precision promised by GPS. More importantly, these tracking sensors and systems are affordable and easy to implement at scale (<a href="https://medium.com/@varun_mani/spatial-data-and-indoor-tracking-6deca00ff1cc"><strong>Spatial Data and indoor tracking</strong></a><strong> </strong>provides a more in-depth look at how indoor tracking systems are enabled by AR/VR).</p><p>In an operational business with frontline workers, spatial data has the potential to augment existing data sources that track business processes (production lines, customer service, inventory etc.) as well as unlock new insights into those business processes. These benefits fall into two broad categories: 1) Process efficiencies that can be gained through the direct capture and analysis of spatial data sources during daily operations and 2) Using spatial data sources to capture, distribute and consume tacit business process knowledge from frontline workers.</p><h3>Process efficiencies</h3><h4>Frontline quality control</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/588/1*HoNwEqAkTmLAAuUQEyj45w.png" /><figcaption><strong>Figure 2:</strong> Simplified process flow diagram of a manufacturing process with common quality checks.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/299/1*FRzDvGYwSFhkVHVp2RF1aA.png" /><figcaption><strong>Figure 3a</strong>: Example process flow diagrams with one quality check at the end of the process vs. four quality checks.</figcaption></figure><p>Frontline quality is usually maintained by a quality control department tasked with building checks into the system such that management can keep a watchful eye on product yields and ensure that the final product or service meets the design specifications. Quality control on a manufacturing line or at a biotech lab today is generally limited to assessing the product or sample at various stages of the process. <strong>Figure</strong> <strong>2</strong> shows a simplified process flow diagram with common quality-checks like IQC (Incoming Quality Check), OQC (Outgoing Quality Check) and IPQCs (In-Process Quality Checks). The primary goal of these QC checks is to ensure there is a feedback loop in the system so that defective products can be removed and the line is working at full-capacity. At its simplest, a line with only one final outgoing quality check (<strong>Figure 3a left</strong>) gives the business a single feedback loop. If the line added three more quality checks (<strong>Figure</strong> <strong>3a right</strong>) then it now has four opportunities to detect and remove defective products as they move through the line. <strong>Figure 3b </strong>helps visualize the efficiency gains and quality improvements that can be had by simply adding more opportunities for QC in the line. Ideally, quality checks occur at infinitesimally small intervals or, in practice, when something happens to the product or any time has passed. When the number of feedback loops are maximized, the “measured yield” at any point will be as close to the actual yield as possible. This figure only visualizes the gains from early detection of defective products; it does not encompass the other benefits of tighter loop quality control checks like bottleneck improvements to reduce the time of each step or manufacturing improvements which would increase the actual target yield.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/597/1*RGnBZM-eTozHfBXeM51J4w.png" /><figcaption><strong>Figure 3b</strong>: Visualization of inefficiencies when changing the number of quality checks in a process by plotting product yield vs. step in production line.</figcaption></figure><p>Unfortunately, there are physical and monetary costs to adding QC checks and every additional QC station extends the line’s throughput time. Most importantly, it has only been possible to place quality checks on the product itself but never on the process a worker is performing. Even the most granular quality control checks have no way of determining if a worker is performing a task correctly as he is doing the task. Today, the state-of-the-art technology to ensure frontline workers in these industries are performing at quality and capacity is Standard Operating Procedures: passive step-by-step documents that workers have been trained on and are supposed to follow at every station.</p><p>What if, instead, it was possible to track every single motion of a human? Where they are on the production line, which step they are on within their station and how they were performing each step? Managers could ensure and verify that their workers were following the SOPs within each station. If it were possible to access the spatial data of each frontline worker combined with a “standard operating spatial data set,” it would be possible to find real time deviations from the SOP at a much smaller granularity than ever before.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/396/1*hhRfsMDlFkc6hE57Z-agVA.png" /><figcaption><strong>Figure 4</strong>: Participant wearing an augmented reality headset so her spatial data can be collected as she performs a simple set of tasks. (source: Waypoint Labs)</figcaption></figure><p>In October 2017, Paul Yarabe and Tatiana Cooke, two students from the Harvard Business School Managing Data Science class and <a href="http://www.waypointlabs.com/">Waypoint Labs</a>, a spatial data and augmented reality startup, put this hypothesis to the test<a href="#_ftn6">[6]</a>. In a controlled experiment, a “standard operating spatial dataset” was created for the simple task of putting various objects into cups at various stations. Participants were then asked to perform the same task while their position and gaze direction were captured in 3-D space (<strong>Figure 4</strong>). These datasets were then compared to determine deviations in the participants’ motion paths from the initial dataset thus creating a continuous quality check for each participant’s actions (<strong>Figure 5</strong>).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/496/1*KDQUN4K8vx5I8ZN7v4Q2zg.png" /><figcaption><strong>Figure 5</strong>: Top-down visualization of spatial data collected as a participant was completing a simple task. The task was to walk to a cup placed at positions A-E, drop an item in the cup, walk back and repeat two more times (A1, A2, A3 and so on).</figcaption></figure><p>In addition to just a quality check, this type of spatial data can also be used to create predictive motion path models to determine which task a participant was doing simply based on their initial motion path and the motion paths of other participants who had done the same task previously. <strong>Table 6</strong> shows<strong> </strong>preliminary data from a small sample size of a machine learning algorithm that was used to predict which cup a participant was moving towards even before they had completed the task. Using spatial data representing only the position of the participants’ heads (without considering where the participants were looking and the orientation of their heads) the algorithm was able to correctly predict human behavior over 70% of the time for certain tasks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/633/1*D5NH79_50prR6c9Hq1RDgw.png" /><figcaption><strong>Table 6:</strong> Preliminary data of a machine learning algorithm trained on a number of participants that was used to predict which cup a new participant was moving towards only based on spatial data.</figcaption></figure><p>The holy grail of quality control is a fully autonomous production line where process variability is so low that an input quality check and an output quality check are all that is necessary. However, there are still more than 2 billion frontline workers doing manual, complex and repetitive tasks daily globally. Capturing and digitizing their motion paths through spatial data provides an unprecedented level of quality control and analysis that can, for the first time, create truly “closed-loop standard operating procedures” in the workplace.</p><h4>Operational efficiency</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*sBLdGM0OsHVVU5Q3yEzQaA.png" /><figcaption><strong>Figure 7</strong>: An example<strong> </strong>of a spaghetti diagram used to re-orient stations in a production facility to improve productivity. Source: <a href="http://slideplayer.com/slide/1585686/">Demystifying Lean Six Sigma — Continuous Process Improvement Overview</a></figcaption></figure><p>Frontline workers do not work in a vacuum. They are constantly interacting with equipment and with other people as they carry out their jobs. Capturing and analyzing the spatial data from these interactions can lead to a more optimal deployment of resources. Lean production methodologies like Six Sigma<a href="#_ftn7">[7]</a> often encourage the creation of spaghetti diagrams to visually map out the physical path a product or a person takes during a process in order to gain insights about operational efficiency. It is very important that these diagrams map out the actual movements through the factory floor or production line instead of the theoretical flow since actual movement paths can vary considerably. Simply moving around stations in a production line to allow a more optimal path for people and products can unlock massive gains in productivity (See <strong>Figure 7</strong>). The more often these diagrams are created and analyzed, the more adaptable the production line is to changes and issues.</p><p>Today, the only accurate way to create these spaghetti diagrams is to have a consulting company or process engineers painstakingly record movements of workers and products as they move through a space. Generally, the space is abstracted down to a lower resolution like stations or rooms to make the movements simpler to record. However, if frontline workers wear augmented reality devices with indoor tracking sensors, this data can be gathered automatically and continuously.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/533/1*Y2CqdA6sVozN1PT1b48f8w.png" /><figcaption><strong>Figure 8a</strong>: A top-down view of a biotech lab with the spatial path data of a lab tech going through a process. The annotated red areas show separate labs while the green line shows the automatically generated path the lab tech took through the process. (source: <a href="http://www.waypointlabs.com/">Waypoint Labs</a>)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/623/1*rxLrMzfshpmH98aZcqWtHQ.png" /><figcaption><strong>Figure 8b</strong>: A 3-d isometric view of the same lab and process.</figcaption></figure><p><strong>Figures 8a</strong> and <strong>8b </strong>show a top down and a 3-D isometric view of a process that was captured using software from Waypoint Labs. This positional data is captured at 60 Hz (60 times a second), is at a resolution of a few inches and includes orientation data so where a person is looking is captured as well. This data is collected from a process that a lab tech must do four to five times every day. From the diagram, we can immediately see that the process uses equipment and supplies from three different rooms and requires the lab tech to walk down a hallway four times to complete the task. Consequently, if the lab collects enough spatial data across all the processes within that lab space, a very strong argument could be made in favor of re-organizing equipment to improve daily productivity. This conclusion could be made with actual, live, data and would cost only a fraction of the current state-of-the-art data collection mechanisms. Furthermore, this data can be played back in real time to recreate the exact actions of the lab tech in full fidelity<a href="#_ftn8">[8]</a>. <strong>Figures 9a </strong>and<strong> 9b </strong>show another example of data that was captured from a different process. In this case, we can see the lab tech spent most of his time at one station in the lab and had to retrieve items from a few locations that were within the same room. Immediately, we can see that this second process is much more efficient to complete than the first one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/470/1*FAuZZTpL3LmoEjPEA829wg.png" /><figcaption><strong>Figure 9a: </strong>Another example of a process captured in a different lab. (Top-down view) (source: <a href="http://www.waypointlabs.com/">Waypoint Labs</a>)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/623/1*EX-Ii4ayFXTvIZY4zFbkgQ.png" /><figcaption><strong>Figure 9b</strong>: A 3-D Isometric view of the same process and lab from <strong>Figure 9a</strong></figcaption></figure><p>Frontline workers, by definition, do not work behind a desk or in static locations. They are constantly moving about a lab, a plant or a production line and interacting with equipment. With spatial data, the nuances and context of all these interactions can finally be analyzed and can lead to more optimal deployment of people, products and physical assets.</p><h4>Safety and Compliance</h4><p>Over quality control and operational efficiency, frontline worker safety is of the utmost importance. Frontline workers have very high-stakes jobs where a mistake here or an oversight there could lead to loss of life or severe injury. Regulatory bodies like OSHA, the FDA, the EPA and many others have spent years developing and honing regulations to prevent injury. Oil &amp; gas workers at a refinery need to follow strict protocols before lighting a pilot light. Workers at a Boeing factory have strict no-step zones on airplane wings and in front of engines to prevent severe injury and damage to the aircraft.</p><p>In nearly all these cases, the physical location and order of tasks outlined in a Standard Operating Procedure is the best guarantee of safety and compliance. Fortunately, the state-of-the-art in spatial data with relation to safety and compliance has already advanced to a great degree. Due to increased autonomy in factories, frontline workers need to work in very close quarters with robots that perform highly dangerous tasks. In many of these cases, light-curtains or radar systems<a href="#_ftn9">[9]</a> are installed to create a virtual spatial fence that can shut off the equipment immediately if a person crosses the threshold. As augmented reality headsets become more common in these industries, the indoor spatial-tracking capabilities of these devices can allow these virtual fences to be reconfigured and scaled at zero additional cost. Workers can be reminded to wear the appropriate personal protective equipment depending on where they are on a factory floor and autonomous robots can even automatically evade humans once they know their locations.</p><h3>Knowledge Transfer</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/279/1*Y3uX7l9Pu-s6iImH2Xe7dA.png" /><figcaption><strong>Figure 10</strong>: Factors contributing to future shortage of skilled production workers based on survey data from executives. Source: <a href="http://www.upskillamerica.org/wp-content/uploads/2015/04/Upskilling_Employer_Handbook_042015.pdf">UpskillAmerica, Deloitte Consulting</a></figcaption></figure><p><strong>Figure 10</strong> shows survey results from executives when asked what the biggest factors contributing to future shortage of skilled production workers are. Baby boomers and economic expansion were the top two contributing factors cited. An estimated 2.7 million jobs are likely to be created due to retirees from the existing workforce and 700,000 jobs created due to natural business growth<a href="#_ftn10">[10]</a>. Even with the private sector spending more than $164 billion each year in education and training<a href="#_ftn11">[11]</a>, the loss of institutional knowledge with older workers leaving and the need to train up a new workforce is one of the biggest problems facing businesses today.</p><p>Desk workers are much further ahead in solving these problems through tools like Dropbox, SharePoint, GitHub and many more. These tools help desk workers quickly codify knowledge and create new processes so that their knowledge is institutionalized, Those same tools then allow them to rapidly distribute this knowledge throughout the company and efficiently consume and modify this knowledge through tools like search and change tracking.</p><p>Frontline workers, on the other hand, find it much more difficult to capture, distribute and consume knowledge at the same ease and pace due to the “desk-less” nature of their work. Apprenticeship, shadowing and standard operating procedures are the best tools available today, but they are all manual and do not scale.</p><h4>Knowledge Capture</h4><p>When a software developer writes new code to perform a task, that code is checked in to a repository that automatically tracks changes, runs tests and requires the developer to document the change. Similarly, when a mechanical engineer creates a new iteration of a part, changes are tracked, adjacent parts are updated, the bill of materials for the final product are automatically generated to include the new part and drawings.</p><p>Both the software developer and the mechanical engineer have the ability to quickly capture and institutionalize the knowledge they create. Conversely, if a biotech lab technician invents a faster way to prepare a sample or if a gas plant worker installs a replacement valve, they need to go through a lengthy and manual process to document the new knowledge they just created.</p><p>With indoor tracking enabled by augmented reality devices, when a lab tech goes to a fridge for some reagent, we can now capture not only which reagent she is using but also which fridge and which shelf she took it from. When she goes over to the centrifuge, we can capture which centrifuge she is using in addition to the speed and time the sample needs to be centrifuged for. Similarly, every step in a Standard Operating Procedure at a gas plant can now be recorded with the precise location of each step and a manufacturing line can finally connect instructions for each station to their physical locations.</p><p>The key difference between the desk worker and the frontline worker is that that frontline work happens in the physical world. This means there are always two components to that work: what the work is and the physical context in which that work happens. Only when the abstract process and the spatial context of that process is captured can the physical process be fully codified. In other words, every time a frontline worker carries out a task or finds a new way of doing something, to fully capture that task, you need to capture what happened and where it happened to fully be able to recreate it.</p><h4>Knowledge Consumption</h4><p>When an accountant is preparing taxes for a client he might want to refer to his client’s previous year’s taxes. A relatively modern accountant probably has digital copies of every tax return he has completed in an organized folder structure by client, year and then by form number. Finding a previous tax return is a trivial process and he can then quickly open it as a pdf or in custom tax preparation software. Since the core asset was fully captured in a digital format, he was able to index and search through that data very efficiently and then consume it in its native digital format as well.</p><p>Once physical processes for frontline workers are captured digitally at full fidelity, similar scenarios can be unlocked for frontline workers as well. A utilities worker who needs to shut down a boiler should be able to simply walk up to the boiler, and find the manuals, maintenance record and any sensor information simply by looking at it. Model numbers, tags and labels in a utilities plant are all crude methods of digitizing physical assets so they can conform to our existing databases and folder structures. In a truly digital plant, a worker should be able to “search” and “retrieve” knowledge by spatial context, or when and where she needs it. Once the worker has shut down the boiler, there should be no need for a separate logging or documentation process. The process of shutting down the boiler should simply be captured as a “new record”. Since we can capture the spatial context as the utilities worker is shutting down the boiler, we can also verify that the task is being completed safely and we now have a spatial audit trail that proves the task was completed to compliance standards.</p><p>Retrieving knowledge is not just limited to operational scenarios like the one above. Once a process has been fully codified, it is in a perfect format for a new employee that needs to be trained on that process. A common training process in frontline industries today is shadowing. Today, if a new employee needs to be trained to perform a task, that employee shadows an expert for a learning period, watching physical motions and absorbing all of the expert’s tacit knowledge. This high-touch model provides a lot of bandwidth between the expert and the trainee but does not scale well past a few trainees. The vast majority of training material today consists of written step-by-step documentation. Businesses that have more capital or processes that require more explanation sometime create multimedia documentation with pictures and videos to help scale training. Detailed video documentation, while expensive, can greatly scale out training since one expert can reach hundreds or even thousands of new trainees. However, this training is inherently linear and cannot adapt to each individual’s familiarity with the process or to the specific version of the process that needs to be learnt.</p><p>If a new trainee can access the right knowledge, at the right time and at the right place, a frontline expert’s knowledge could truly be scaled out transitioning the inefficient 1:1 shadowing model to a 1:many model. Capturing knowledge from an expert as described above, allows us to create a digital representation, or avatar, of the expert as they are going through a process. If this avatar can fully mimic the actions of the expert, a trainee could simply shadow this digital recreation and have a much higher bandwidth experience. While this scenario is still only a one-way transfer of knowledge, the knowledge from one expert can be transferred at a much higher fidelity than before because of the spatial context associated with that data.</p><h4>Knowledge Distribution</h4><p>Capturing and consuming knowledge with a layer of spatial context finally gives frontline workers access to some of the most fundamental tools of the trade that desk-workers enjoy. Once the tools available to frontline workers for capturing and consuming knowledge mature enough to match the speed and frequency of those available to desk-workers, the real potential of spatial data can be harnessed.</p><p>GitHub revolutionized the way software developers shared, re-used and published code. Similarly, a frontline worker that finds a better way of doing a process should be able to send a “pull request”<a href="#_ftn12">[12]</a> for a physical process with a modification to that process. Once that pull request is approved by a process engineer, the process can be updated with the improvement, and that new process can then be pushed to every frontline worker across the globe while the older version is archived.</p><p>In the late 1900s, Toyota pioneered the concept of a worker pulling an Andon cord<a href="#_ftn13">[13]</a> so that the production line could prioritize a process improvement. Spatial data can bring that concept to the 21st century and help companies realize a new standard in continuous improvement.</p><h3>The Future of Spatial Data</h3><p>We are simply scratching the surface of spatial data today. The concept of tracking humans in indoor space has a long way to go before we can unlock its full potential. Today, augmented reality devices and the associated algorithms and sensors allow us to track the position and orientation of a person’s head with great precision. Soon, eye-tracking will enable us to record the location of a person’s attention, hand-tracking will allow us to record every motion of a person’s fingers and precision tracking devices will allow us to track fine-grain movements like a surgeon’s scalpel. Just as advancements in processing power, storage and connectivity improved productivity for desk workers, each advancement in spatial tracking will unlock new scenarios and enrich existing scenarios for frontline workers.</p><h3>Endnote: Privacy Considerations</h3><p>Spatial data is a concept that has existed for decades but until recently, it has been limited to low-resolution, outdoor data. Even at this level, location data has been a very sensitive topic rife with controversy because people are inherently sensitive about the types of information corporations and the government are able to collect. Uber has come under fire for tracking user’s locations without their consent<a href="#_ftn14">[14]</a> and Google Street View has had to face numerous controversies<a href="#_ftn15">[15]</a>. The benefits of tracking objects at a much higher precision and indoors described in this paper needs to be coupled with similar considerations from a privacy perspective as well. The way spatial data is stored and anonymized needs to be regulated so that individuals maintain a certain level of control over what can and cannot be tracked. Similarly, it is still unclear who owns an individual’s spatial data if they are working a shift on a production line or factory floor. Like any new technology, each new benefit is accompanied by a litany of possible missteps and it is the responsibility of those who make those technological advances to constantly review it in the context of human society.</p><p><a href="#_ftnref1">[1]</a> <a href="https://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/the-case-for-digital-reinvention">https://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/the-case-for-digital-reinvention</a></p><p><a href="#_ftnref2">[2]</a> <a href="https://blogs.windows.com/devices/2017/11/01/mixed-reality-momentum-continues-modern-workplace-microsoft-hololens-expands-29-new-markets/">https://blogs.windows.com/devices/2017/11/01/mixed-reality-momentum-continues-modern-workplace-microsoft-hololens-expands-29-new-markets/</a></p><p><a href="#_ftnref3">[3]</a> <a href="https://insights.staffbase.com/blog/non-desk-frontline-offline-employees-internal-communications-for-all-infographic">https://insights.staffbase.com/blog/non-desk-frontline-offline-employees-internal-communications-for-all-infographic</a></p><p><a href="#_ftnref4">[4]</a> <a href="https://www.gartner.com/smarterwithgartner/gartners-top-10-technology-trends-2017/">https://www.gartner.com/smarterwithgartner/gartners-top-10-technology-trends-2017/</a></p><p><a href="#_ftnref5">[5]</a> <a href="https://hbr.org/2015/10/how-smart-connected-products-are-transforming-companies">https://hbr.org/2015/10/how-smart-connected-products-are-transforming-companies</a></p><p><a href="#_ftnref6">[6]</a> Tatiana Cooke &amp; Paul Yarabe, The Wrong Way to Seek a Goldfish: Predicting Motion paths using Augmented Reality Technology</p><p><a href="#_ftnref7">[7]</a> <a href="http://www.six-sigma-material.com/Spaghetti-Diagram.html">http://www.six-sigma-material.com/Spaghetti-Diagram.html</a></p><p><a href="#_ftnref8">[8]</a> <a href="https://www.youtube.com/watch?v=rQgbZokCSEc">https://www.youtube.com/watch?v=rQgbZokCSEc</a> shows an example of a short process that was captured with Waypoint Labs technology.</p><p><a href="#_ftnref9">[9]</a> <a href="https://www.technologyreview.com/s/608863/a-radar-for-industrial-robots-may-guide-collaboration-with-humans/">https://www.technologyreview.com/s/608863/a-radar-for-industrial-robots-may-guide-collaboration-with-humans/</a></p><p><a href="#_ftnref10">[10]</a> UpSkill America, Deloitte Consulting, <a href="http://www.upskillamerica.org/upskilling-employerhandbook/">http://www.upskillamerica.org/upskilling-employerhandbook/</a></p><p><a href="#_ftnref11">[11]</a> Business Roundtable, Work in Progress: How CEOs are helping close America’s skills gap: <a href="http://businessroundtable.org/skills">http://businessroundtable.org/skills</a></p><p><a href="#_ftnref12">[12]</a> <a href="https://help.github.com/articles/about-pull-requests/">https://help.github.com/articles/about-pull-requests/</a></p><p><a href="#_ftnref13">[13]</a> Liker, Jeffrey (2004) “The Toyota Way”</p><p><a href="#_ftnref14">[14]</a> <a href="https://www.theverge.com/2017/8/29/16219542/uber-location-tracking-app-ios-android-privacy">https://www.theverge.com/2017/8/29/16219542/uber-location-tracking-app-ios-android-privacy</a></p><p><a href="#_ftnref15">[15]</a> <a href="https://en.wikipedia.org/wiki/Google_Street_View_privacy_concerns">https://en.wikipedia.org/wiki/Google_Street_View_privacy_concerns</a></p><h3>Acknowledgements</h3><p>- Prof. Joseph Fuller, Harvard Business School for advising me through this Independent Project</p><p>- Michael Elliott, Hans Gertje, Laura Maciejko, Allen Thaya Kumar and Baraneedharan Ulaganathan, Mitra Biotech Ltd. for providing countless hours of testing time in the lab and feedback while we tested our prototypes</p><p>- Tatiana Cooke and Paul Yarabe, Harvard Business School for co-running the spatial data experiment outlined in the paper and analyzing the results</p><p>- Sara Remsen and Umar Arshad, Waypoint Labs, Inc. for help building and testing the software prototypes, editing this paper and for being the most amazing co-founders in the world.</p><p>- The Martin Trust Center for Entrepreneurship and the Delta V accelerator, MIT, for providing space and advice to help us build our product and company</p><p>- The Harvard Innovation Lab &amp; Rock Accelerator, Harvard Business School, for providing space and advice to help us build our product and company</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c77e6cb19504" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Spatial Data and indoor tracking]]></title>
            <link>https://medium.com/@varun_mani/spatial-data-and-indoor-tracking-6deca00ff1cc?source=rss-29643553e9------2</link>
            <guid isPermaLink="false">https://medium.com/p/6deca00ff1cc</guid>
            <category><![CDATA[virtual-reality]]></category>
            <category><![CDATA[spatialdata]]></category>
            <category><![CDATA[augmented-reality]]></category>
            <dc:creator><![CDATA[Varun Mani]]></dc:creator>
            <pubDate>Mon, 09 Jul 2018 21:16:08 GMT</pubDate>
            <atom:updated>2018-07-09T21:49:47.770Z</atom:updated>
            <content:encoded><![CDATA[<h3>Core components of Spatial Data</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/284/1*60LE9B-_dQgBptAoOEio9A.png" /><figcaption><strong>Figure 1</strong>: Cartesian representation of a common origin and offsets in the X, Y and Z axes to represent a 3-D position in space.</figcaption></figure><p>At its core, a spatial position system needs to provide two things: first, a common understanding of an origin and second, an offset from that origin in the three spatial dimensions. (see graphic) In GPS, the common origin is the earth itself and the offsets are described as latitude, longitude and altitude. In contrast, indoor spatial positioning systems can have an arbitrary origin that the objects can be positioned relative to. For example, Bluetooth beacons are used in some malls to track visitors. In this case, each beacon acts as an origin and can track the location of individual’s mobile phones. While single beacon tracking is useful, signals from multiple beacons can be combined, along with a commonly agreed upon origin, to create much more sophisticated indoor positioning systems<a href="#_ftn1">[1]</a>. Instead of Bluetooth beacons, one can use WiFi, Infrared or even cameras with computer vision algorithms to track objects spatially.</p><h3>Tracking methods</h3><p>There are two methods to track objects in 3-D space, namely <em>Outside-In</em> tracking and <em>Inside-Out</em> tracking.</p><p>The methods described in the previous section, and any other method which uses some sort of beacon, all fall into the category known as Outside-In<em> </em>tracking. In this method, some object (a beacon, tower or camera) is placed in an environment and then either sends or receives signals to the objects it needs to track<a href="#_ftn2">[2]</a>. Depending on the signal, the object being tracked and the beacon, various algorithms (signal strength, time between signals, object size etc.) are then used to triangulate the position of the object to varying degrees of precision. Many Virtual Reality systems that are currently in the market utilize different forms of Outside-In tracking. The HTC Vive, for example, uses external towers and infrared emitters to precisely track the location of the headset. The Oculus Rift and Playstation VR use an external camera instead to track the location of the headset.</p><p>Inside-Out tracking, conversely, does not require an external sensors or beacons. Instead, the object being tracked has one or more cameras and an inertial measurement unit (magnetometer, gyroscope and accelerometers) on it which looks out at the environment and senses it. Sophisticated computer vision algorithms like 3-D Simultaneous Localization and Mapping (SLAM) algorithms are then used to get the precise location of the object in that space. Inside-Out tracking tries to mimic a human’s ability to remember locations from memory and navigate through a space using just her eyes and inner-ear.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/635/1*FdExYsAif0ycixOuxaYdWQ.png" /><figcaption><strong>Figure 2</strong>: Illustration of Outside-In vs. Inside-Out tracking</figcaption></figure><p>While Inside-out tracking does not require the external sensors that Outside-In tracking requires, the lack of external sensors means that there is no obvious common origin that can be chosen. In an Outside-In tracking system, the external sensors do not (and cannot) move since the underlying algorithms assume that the external beacons are static. This inherent attribute of Outside-In tracking means that the origin is commonly placed in the center of all the beacons to give the most precise results. For Inside-Out tracking, since there is no existing notion of an origin, the initial location of the object once it is tracked is usually used as the origin. <strong>Figure 2 </strong>shows an illustration of Outside-In vs. Inside-Out tracking and <strong>Table 3</strong> enumerates some of the differences between the two tracking methods.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/624/1*HPd1uZPYcBSKrXJjTDjfhA.png" /><figcaption><strong>Table 3</strong>: Comparison between Outside-In vs. Inside-Out tracking</figcaption></figure><h3>Indoor vs Outdoor tracking</h3><p>We have been leveraging spatial data since the dawn of the Global Positioning System. GPS is an example of outside-in tracking where geostationary satellites (static “beacons”) generate radio signals so that individual objects can be tracked relative to the position of the earth (the common origin).</p><h3>Outdoor GPS tracking</h3><p>Getting an accurate location in 3D space of an object helps us navigate to work every day, helps a fleet of UPS trucks report the location back to headquarters and even helps guide missiles. GPS also allows airline and shipping companies to optimize routes and get a snapshot of the exact location of all their assets. Uber uses the spatial data from GPS to incentivize drivers to move to areas with higher demand. Location-aware apps on our mobile phones have enabled brand new consumer and retail scenarios to target customers based on their GPS location.</p><p>Unfortunately, this precise and ubiquitous tracking technology is rendered useless as soon as one goes indoors. Since GPS is based on receiving signals from a satellite, devices need to fallback to less precise spatial tracking mechanisms like Bluetooth, or WiFi.</p><h3>Indoor tracking with Augmented and Virtual Reality</h3><p>Fortunately, with the advent of augmented and virtual reality, indoor tracking has become much more commonplace and at a price range that is much more accessible and easy to install. AR and VR devices need to track the location and orientation of the user very precisely. The only way to trick people into thinking a virtual object is static is to orient virtual objects perfectly based on the perspective and movements of the user. In other words, if the person moves one foot to the left, the device needs to detect that and then move everything in the displays one foot to the right so that it looks like nothing moved. The more quickly and accurately the device can do that, the better the illusion is to the person wearing the device.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/624/1*tdJe3MYYOqsvdE9i84o0gw.png" /><figcaption><strong>Figure 4</strong>: The six degrees of freedom to fully capture an object’s position and rotation in 3-D space. The six degrees are: Forward, Back, Left, Right, Up, Down, Yaw, Pitch and Roll.</figcaption></figure><p>Early VR devices did not have a way to track a user’s location in 3D space. With simpler IMUs (inertial measurement units) these devices could only determine the orientation of a user (yaw, pitch and roll). This allowed static VR experiences where a user could look up or down and left or right but could not move around in the space. To fully capture an object’s rotation and position in 3-D space, all six degrees of freedom need to be captured (<strong>Figure 4</strong>). More recent VR and AR devices have been able to accomplish this through Outside-In tracking but required prior installation of static beacons, towers or cameras in the environment. With the advent of more advanced AR and VR devices, more precise sensors and Inside-out tracking algorithms, we are finally able to capture all six degrees of freedom in a form-factor that requires no prior installation and an unprecedented level of precision.</p><p>See: <a href="https://medium.com/@varun_mani/spatial-data-the-next-frontier-of-business-intelligence-c77e6cb19504">Spatial Data: The next frontier of business intelligence</a> for how this kind of data can be used to change the way your business runs.</p><p><a href="#_ftnref1">[1]</a> Filip Mazan and Alena Kovarova, A Study of Devising Neural Network Based Indoor Localization Using Beacons: First Results <a href="http://www2.fiit.stuba.sk/~kovarova/files/publications/CISJ%20Vol19%20No1%202015%20Mazan.pdf">http://www2.fiit.stuba.sk/~kovarova/files/publications/CISJ%20Vol19%20No1%202015%20Mazan.pdf</a></p><p><a href="#_ftnref2">[2]</a> Manikanta Kotaru and Sachin Katti, Position Tracking for Virtual Reality Using Commodity WiFi <a href="http://openaccess.thecvf.com/content_cvpr_2017/papers/Kotaru_Position_Tracking_for_CVPR_2017_paper.pdf">http://openaccess.thecvf.com/content_cvpr_2017/papers/Kotaru_Position_Tracking_for_CVPR_2017_paper.pdf</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6deca00ff1cc" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>