<?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 Diego Botin on Medium]]></title>
        <description><![CDATA[Stories by Diego Botin on Medium]]></description>
        <link>https://medium.com/@diegobosa?source=rss-e703c3132b4a------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*DmJfEalIv-uWsYd3koYaVw.jpeg</url>
            <title>Stories by Diego Botin on Medium</title>
            <link>https://medium.com/@diegobosa?source=rss-e703c3132b4a------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 08:03:51 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@diegobosa/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[The Role of AI and Modern Technologies in Motorsport]]></title>
            <link>https://medium.com/@diegobosa/the-role-of-ai-and-modern-technologies-in-motorsport-6ae23c84dfe6?source=rss-e703c3132b4a------2</link>
            <guid isPermaLink="false">https://medium.com/p/6ae23c84dfe6</guid>
            <category><![CDATA[motorsport]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[future]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[formula-1]]></category>
            <dc:creator><![CDATA[Diego Botin]]></dc:creator>
            <pubDate>Sun, 20 Sep 2026 08:01:58 GMT</pubDate>
            <atom:updated>2026-09-20T12:27:11.166Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hAg8VYUIQslXhK6EaMTtbQ.jpeg" /><figcaption>Author generated with AI.</figcaption></figure><p>In a recent panel, I got asked about my experience and thoughts on the role of AI and modern technologies (such as Digital Twins, Machine Learning, IoT, etc.) in the Motorsport industry. And with all the recent talk about AI’s possibility of killing all humanity, solving millenium mathematical problems, and hacking systems, I felt it was a good time to talk about this and hopefully open up the discussion.</p><p>Honestly there is a very simple (boring) yet generalistic answer to the question: AI will be embedded in all industries in the near future and motorsports are not the exception. Motorsport is on the verge of adopting AI at scale, but meaningful integration will only happen when these modern tools become trustworthy, physics‑aware, and aligned with real engineering problems rather than hype.</p><p>Now, the motorsport industry works and behaves <strong>VERY </strong>differently to the automotive industry (topic for future discussions). Different project lifecycles, different team organization, objectives, regulations and standards, etc. So, here are some of my deeper points about these new technologies which hopefully sparks some thought:</p><h4><strong>AI For Engineering</strong></h4><p>I’m sure everyone is using by now chatbots (such as ChatGPT, Copilot, Claude, etc.) for searches, questions and even some engineering help. Some are using agents to automate simple tasks such as calendar organization, email replies, meeting notes, debugging, etc. But what about real, integrated engineering work? Turns out that integrating new technologies into a team’s workflow can actually be highly disruptive and might look like an impossible software landscape to traverse. One of the great obstacles for implementing new technologies is the ignorance of the capabilities and possibilities, but also the unwillingness of changing things to adapt to new workflows and processes. And so, this raises questions such as what can we use AI for? How can we trust its results? How do we safely give it access to confidential information without having a security vulnerability? What will we do if someday it stops working (what’s the plan B)? And maybe more importantly: what problem can it solve which is so valuable that we are willing to invest on it with an already capped-budget?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*GwHT1a-ZouPT3gsS.jpg" /><figcaption>Logos for the most popular AI chatbots.</figcaption></figure><p>As is widely known, Formula 1 teams need to adhere to a cost cap appointed by the FIA which determines how much teams can spend in any given year. So, for teams to spend great amounts on a project, they need to be sure that its results will bring enough value to apoint money to it. And its not just about buying software and licenses but often, the full integration of new technologies requires hiring new personnel, updating current hardware, rethinking processes, and exposing vulnerabilities and issues.</p><p>So, why would they spend time and money on developing or integrating AI systems if they don’t have a clear problem to solve? I remember asking a high-rank engineer in one of the big teams about his take on implementing AI to their workflow. He answered that certainly AI sounds attractive, but he’s looking for solutions to their problems, not problems for solutions. Meaning AI presents a great solution for many problems, but it doesn’t make sense to implement a new technology just because it <strong>could </strong>solve problems. Ideally, you have a problem you need to solve and then look for a way of solving it, not the other way around. So, if there is no immediate need for AI, why implement it?</p><p>There are many companies and public initiatives building data-driven models (for tyres or aerodynamics for example), building agents to perform engineering processes, or systems that attempt to find patterns in data and potential improvement opportunities. But the question is: the problems that this new systems are solving, are they the problems that the industry actually needs solving with AI? <strong>Or are we just solving them with AI because it’s trendy?</strong> Don’t get me wrong, this technology is very powerful and capable, but it may not be the solution that teams need for the current problems. We see these technologies used commonly in applications such as fuel load estimation, agent-based game-theoretic strategy simulation, setup optimisation, anomaly detection, product lifecycle management, predictive maintenance, etc. but these are individual and isolated applications. Can we implement a platform based on AI that integrates all these applications?</p><h4>The problem of confidence</h4><p>In an industry where everything evolves and moves so fast (not just the cars), engineers need tools which they understand and can be fully confident about the results. We already use simulations (as I talked about in a previous blog — read <a href="https://medium.com/@diegobosa/understanding-racing-through-simulation-part-1-why-simulation-and-different-types-of-simulation-a9af3fcd5ae0?sharedUserId=diegobosa">here</a>) to predict the future, to understand systems, and to take design decisions (which is sometimes refered to as model-based design). So obviously, we need simulations we can trust to be right. And if they’re not, we need to be confident that we understand how they work so we can fix them, or expand them in the future. Think about this: in a race, strategy engineers might have a few seconds to make the call for a pit stop or not, and the specification of the pit stop. They will be running Monte Carlo simulations, data driven simulations and many more calculations during the weekend to build up the knowledge. So, in this time-pressured scenario, there is no time to run a simulation which you don’t fully understand and don’t trust 100%. Sure, “all simulations are wrong”, but it’s a matter of using the simulation which is least wrong and more consistent.</p><p>Let’s be clear, AI and ML have been used in the industry before, it’s not a completely new technology, but I don’t think it’s part of the day to day workflow and operations of engineering yet. Maybe in specific applications, but not as a widely adopted tool.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZUiXMW38KIfBX-9QeRp43A.jpeg" /><figcaption>Author generated with AI.</figcaption></figure><p>The problem with most AI/ML models is that they behave as black boxes. We have this really complex and powerful models that are trained on great amounts of data. And so they end up being statistical models that predict what will happen based on what they were trained on. There is no physical meaning to the results here, that’s why they are called data-driven models. This doesn’t mean that they are not powerful, they bring great value to the table and end up being really good and fast at predicting dynamic systems within their trained dataset. They can also bring greater understanding about underlying interactions between parameters and variables and I think more importantly, they can evolve with time. They can continue to be trained, improved and made more robust. In a Digital Twin architecture, they can be used to continue to represent a physical system (or physical twin) even as the physical asset evolves and updates.</p><p>Recently, there is a big push on AI that does not behave just as black boxes. Modern methodologies such as physics-informed neural networks (PINNs), cyberphysical models, and explainable AI (xAI) really try to bring out the most out of AI whilst still holding some physical and explainable meaning to the models (more on this in future blogs). Meaning, they enable engineers to use the data-driven approach, but also be constrained by physics such that the models can predict outside of the training dataset, and they can also have accurate and precise results even with a small amount of data to train it on. I think this is the technology that will really bring AI/ML to engineering (at least in motorsports). Mix that with agents and you’ve got a proper AI workflow going!</p><h4>The confidentiality boundary</h4><p>It is well known that motorsports (especially at high stakes such as in Formula 1), there is a lot of secrecy and confidentiality concerns. Teams build in silence and win in public. So unfortunately for academics, there is not a lot of public information about how teams work. In fact, Formula 1 cars are so unique, that when asking an old boss of mine about how people can learn more about Formula 1 engineering to become an F1 engineer, he said that you can only learn F1 engineering by being an F1 engineer. You can learn the fundamentals (which are <strong>so </strong>important to get right) such as vehicle dynamics, programming, modelling, simulation, aerodynamics, etc. But F1 gearboxes are so unique that you will only learn about it by being in F1. Embedded software for an F1 car is so unique that you will only understand it by actually working in F1. So, there is a “bespokeness” to it all that means that universal or common models and systems are not readily applicable in F1. The actual car behaves differently is many areas as commercial cars and this, coupled with bespoke control strategies and bespoke software solutions is what makes F1 cars so unique.</p><p>You do see some really good academic articles and MSc/PhD projects where they work on motorsports and more commonly Formula 1, Formula E and WEC applications. Doing research for a literature review on the topic I found that there is particular interest around energy management, data-driven models for physical systems and minimum laptime problems. However, we are yet to see AI driven <strong>workflows</strong>…</p><p>Now, motorsports already have a big advantage that many automotive companies don’t. For example:</p><ul><li><strong>The availability of real time and historical data from cars and races.</strong> Years and years, hours and hours of running with traceable information which enables training models with a big dataset. This of course enabled by hundreds of sensors on the cars and a proper telemetry and data management infrastructure.</li><li><strong>The predictability of drive cycles and stress on the car. </strong>We literally know where the car will be driven, the trajectory, how the driver actually drives it, and so much information about the car itself. It also helps that F1 drivers can be so consistent that you have many many laps with very consistent driver inputs. This predictability makes anomaly detection and reinforcement learning easier to implement.</li><li><strong>The relatively flexible space to innovate and implement new technology</strong> (especially in Formula E for the software side of things). Although we are regulated by the FIA, there is still room (in some areas) for lots of innovation — especially in the operational side.</li></ul><p>So, motorsports I think presents a very interesting playing ground to implement and develop new technology. I think we are going to see a very interesting near future where AI will become a part of racing in many aspects and while some think it will negatively impact motorsport, I think from the engineering side, it will be an exciting time. We need to be careful as motorsports are in the end, a sport. And the human side must never disappear. This is what makes racing exciting most of the times.</p><h3>Closing thoughts</h3><p>So where do I think the Motorsport industry is in terms of full adoption and integration of AI and modern technologies into their workflows?</p><p>I think motorsport and Formula 1 in particular is in a really interesting and exciting time where these technologies are not necessarily fully adopted and integrated into teams’ operations, but we are getting close to a breakthrough. A breakthrough where a team is going to gain exceptional competitive advantage from using these tools correctly and as soon as that happens, every team is going to jump on the bandwagon. And by breakthrough I don’t mean a specific application of AI or Digital Twins, I mean a full adoption and integration of these technologies in how the team operates. Think about full AI-augmented simulation pipelines, Digital-Twin-based car development, automated setup optimisation and evaluation, real-time decision support systems based on AI/ML/Digital Twin technology, etc.</p><p>We see a lot of teams joining forces with AI companies (e.g. Williams with Claude, Aston Martin with Cognizant, etc). So inevitably, it’s a matter of time. And this adoption will undoubtedly bring many changes to how teams operate, to regulations and policing efforts, and more importantly <strong>how us engineers work, and how we bring value to the industry</strong>. We must re-invent ourselves and how we work.</p><p>My bachelours degree is in mechatronics engineering and when I started studying this career, it was relatively new. The main objective for mechatronics engineering is automation. Automating production lines with robots, designing robots for tasks to replace humans, etc. So, naturally there was a lot of worry around how mechatronics was going to take the jobs of so many people. However, it is important to look at the grand scheme of things.</p><p>Think about the Industrial Revolution. People had jobs like weaving things by hand, waking up other people at a specific time and plowing fields by hand. And all of that got replaced with machinery. Yes a lot of people lost jobs, but new ones got created. And I think this is how we need to look at the new digital era with AI. We must find what makes humans valuable and good at things and we must also adapt to new technologies. There is a responsibility from employers to make sure that employees get to develop beyond the tasks that can be automated and I think this gains great relevance during modern times. Where a technician might be replaced in a production line with a robot, we must ensure that the technician gets to develop such that they then have the capability of operating that robot, or designing that production line, or at least maintaining and servicing those robots.</p><p>Sure, there is a lot of uncertainty about what the future holds. But isn’t that exciting?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*TdkSxCfObtJRwKxh.png" /><figcaption>Author generated with AI.</figcaption></figure><h3>Disclaimer</h3><p>I am a engineering in the field so this is me sharing with you some of my personal experience over the years in academia/research, personal projects, and professional experience.</p><p><em>This article is unofficial and is not associated in any way with Formula 1, or any related companies and/or teams. The comments expressed in this article and in the analysis are personal and do not represent the position of any company and/or team. This article is for fan use and educational use. The content and materials are intended solely for educational and informational purposes.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6ae23c84dfe6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Understanding Racing Through Simulation. Part 1: why simulation and different types of simulation]]></title>
            <link>https://medium.com/@diegobosa/understanding-racing-through-simulation-part-1-why-simulation-and-different-types-of-simulation-a9af3fcd5ae0?source=rss-e703c3132b4a------2</link>
            <guid isPermaLink="false">https://medium.com/p/a9af3fcd5ae0</guid>
            <category><![CDATA[engineering]]></category>
            <category><![CDATA[racing]]></category>
            <category><![CDATA[motorsport]]></category>
            <category><![CDATA[formula-1]]></category>
            <category><![CDATA[simulation]]></category>
            <dc:creator><![CDATA[Diego Botin]]></dc:creator>
            <pubDate>Wed, 21 Jan 2026 20:03:03 GMT</pubDate>
            <atom:updated>2026-01-21T20:03:03.787Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/950/0*HcZfO6UgE43FMsjI.jpg" /><figcaption>Figure 1. DMG-1 Motorsport Simulator from Dynisma [1].</figcaption></figure><h3>Introduction</h3><p>Formula 1 and motorsports in general are an expensive sport. Not only for drivers and their families, but for car development itself and even tickets to races! Teams spend millions on development costs to try and get the car to a state where it is FAST and not only fast, but THE fastest on the grid. And sometimes there will be some designs and concepts taken over from year to year but the development will involve integrating what the team learned from last season into a new car. Additionally, regulation changes present the opportunity to drastically change up the overall concept and philosophy of the car and start from “scratch” (such as the F1 2026 regulation changes which will see a completely new powerunit, changes to the aerodynamic components and their operation, changes to the ECU and telemetry network, and dimensional changes to the car ). So, how do F1 teams create a car, powerunit, or setup from scratch? The answer is simulation.</p><p>We live in an F1 era where physical testing of cars is regulated and capped. This means that teams are only allowed to test their cars on the track at specific events regulated by the FIA, and the time they are allowed to use wind tunnels for aero development is also limited to a mandated number of hours. Furthermore, there is a budget cap, so teams are allowed to spend a maximum of ~$135 million a year [2] including developments costs, operations, logistics, etc. However, teams clearly need to test new designs, new setups, new strategies, etc. So here is where simulation comes in.</p><p>There is a plethora of simulation techniques and algorithms that can and are used in engineering a race car, I will briefly go over some of them, their applications and their benefits. After all, engineers may use some external off-the-shelf simulation environments and software, but they will also develop their bespoke simulation environments and models. In the same way they do with the actual race car, engineers push simulation technology to the limit with the objective of having the most realistic and robust simulation that most accurately represents what the real race car is or will experience on the track. But… <strong>why</strong>?</p><h3>Why use simulation?</h3><p>Put quite simply, the end goal for engineers is to understand the car as deeply as possible and as accurately as possible in order to continuously improve it. And simulation provides an relatively cheap environment (compared to track testing) where engineers can test and experiment without: a) spending so much resource (time, money and using up hardware), and b) potentially putting a driver at risk. As a reference, for a team to run an F1 car on track, it may take a big crew of anywhere between 10–30 people at least (think PU engineer, performance engineer, race engineer, tyre engineer, IT engineer, systems engineer, logistics and operations, mechanics, driver, etc.). However for a team to run a Driver-in-the-Loop simulation, it may take a crew of maybe 2–3 (driver, and 1–2 simulator engineers), half the time and almost no logistic operations other than preparing a run plan.</p><p>Furthermore, offline simulation technology allows individual engineers to test aerodynamic concepts, car setups, reliability experiments, etc. So, with a proper simulation environment, teams can test faster, cheaper and therefore iterate their designs a lot faster. Ever wonder how F1 teams bring big updates to their cars within a few weeks across the world?</p><p>So, the overall benefits of using simulations are clear. However, what are some specific ways that simulations helps to maximize the competitiveness of the team? Here are some examples:</p><ol><li>- <strong>To be able to understand the car in a variety of what-if scenarios such as different ride heights, balance configurations, aerodynamic packages, energy deployment strategies and even race strategies</strong>. Simulations allow engineers to evaluate an inmense amount of configuration combinations and create a performance envelope for the car (i.e. know what are the performance limits of the car).</li><li><strong>Understanding sensitivity to changes.</strong> An example is understanding the lap time sensitivity to ride height. So, multiple simulations can be ran at different ride heights, and using lap time as a performance metric, engineers can understand: a) which ride height is best to minimize lap time and b) how an increase or decrease in ride height will ultimately affect lap time. Of course lap time is not the only metric as it may not represent drivability or an adequate race-distance pace.</li><li><strong>Creating the best setups for each race and/or event. </strong>As mentioned in the previous point, not necessarily the best setup for a lap is the best setup for a whole race (think about tyre degradation, battery degradation, etc.). So understanding what is best for a whole race length is important.</li><li><strong>Developing future control strategies and testing future concepts (vehicle, systems, packages, etc.).</strong> This is a major advantage of simulations in a model-based design philosophy. One can test a new energy deployment strategy without having to test it live on a race where the stakes are higher. One could also test a completely new suspension design without having to even design it on CAD.</li><li><strong>Find winning strategies to use during specific races.</strong> And this goes beyond the setup or design of the car itself. Of course one needs to account for the pace of the car and the other cars on the grid, but this is more about finding the best race strategy. So, answering questions such as: in what lap should we pit?, how would a safety car affect our decision of when to pit (having a plan A, plan B,…. plan Z), what tyres should we fit when we do pit? and even analysing your team’s competitiveness against other teams.</li></ol><p>More importantly, and to summarize the answer to the question of why simulation: <strong>simulation enables engineers to take data-drive decisions.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Yk3-PITesPJGaeiQJayPbQ.png" /><figcaption>Figure 2. AI generated image of engineers taking data-driven decisions in the control room.</figcaption></figure><h3>Different types of simulations, different objectives, same purpose</h3><p>So, let’s dive into different types of simulations, how they work (broadly) and what they can be used for.</p><h4>Lap-time simulation</h4><p>A lap-time simulation (LTS), as its name implies, is a simulation of a single lap to evaluate lap-time. In essence, the objective is to be able to evaluate the performance of the car for a specific track. Creativity and time is the limit, but this tool allows engineers to perform sensitivity analysis, test a few car setups to determine the starting setup for an event, evaluate the stability of the car in a track, etc. And the complexity and accuracy of the simulation is down to the engineers. They can play with different types of vehicle models and system models, track models, and solving algorithms.</p><p>In terms of vehicle modelling, the main top models can be point-mass models, single track or bicycle models, and full vehicle models. In the end all models are simplifications of the real car, but as one adds the four tyres, suspension compliance, aerodynamics, etc. into the equation, the accuracy of the simulation will improve — at the cost of increased computational burden. Furthermore, in each of the models I have mentioned, one could add more system models. For example, adding different tyre models, powertrain models, aerodynamics models, driver models, etc. Note that the fidelity of these individual models can be a mixture of low and high fidelities. For example, having a really complex powertrain/gearbox/transmission model but a very simple or fixed aerodynamic model. Similarly, track models can be as simple as X and Y coordinates, or as complex as 3D ribbon models [3].</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1N0_UBOiTRFX7ctxnImiAw.png" /><figcaption>Figure 3. Vehicle model types. Point mass model (left), single track model (middle) and four wheel model (right).</figcaption></figure><p>Finally, in terms of the solving algorithms, there are multiple techniques that can be used to solve the models stated above. And both the solving technique and modelling technique are intrinsically related and dependent on the application. So there may not always be a need for the utmost accuracy and fidelity of the model. In some cases, a more simple simulation can be levaraged to allow rapid tests and testing changes to the car that may not have a huge impact on performance. It all depends to what the engineers want to achieve with that simulation. Anyway, examples of the most common techniques are:</p><ul><li><strong>Quasi-steady state simulation: </strong>where the car is assumed to always be in steady state (not realistic as the car is always moving and there is always transient behaviour) but solved at short time intervals. This way, the resulting behaviour looks like a transient simulation, but is inherently not capturing transient and dynamic behaviour. To explore an open source example, I recommend OpenLap in Matlab [4].</li><li><strong>Transient/dynamic simulation:</strong> in this simulation, the car is not assumed to instantaneously be in steady state, but rather systems evolve through time and are evaluated through time. This way, other phenomena like momentum and inertia can be part of the evolving model results.</li><li><strong>Hybrid approach: </strong>This technique involves using a QSS solver but simulating some modules or subsystems in a transient way. Of course this cannot happen at the same time, but an approach may be to solve the QSS solution first, and then calculate signals or measurements that actually evolve through time. QSS plus transient system models.</li></ul><p>As an open question: What are some applications where you have used or would think that using LTS is valuable?</p><h4>Optimisation</h4><p>Optimisation is a bit different than what we learned in lap time simulation. In LTS, you are doing an evaluation of a vehicle model to find the performance envelope and then calculate how quick a car can go through a lap in a given path and with a given setup. With optimisation however, you can find the best compromises to minimize lap time. In a minimum lap time problem for example, you may choose to optimize for the best race line to minimize lap time. Or you may want to find what is the weight distribution, energy deployment strategy, gear shift strategy or brake balance that will minimize lap time.</p><p>More complex implementations can have multiple objectives optimised at once or in series. For example, finding the best energy deployment strategy might need to optimize for the optimal use of gear shifts first or at the same time. Or one can design the optimal control problem (OCP) to optimise for the minimum lap time with the minimum amount of energy used, or to distribute a fixed amount of energy in the lap to minimize lap time (which is the typical problem in Formula 1’s hybrid powertrain era).</p><p>This flexibility of finding optimal compromises for multiple (and often conflicting objectives) is great for engineering design, but it does come with a cost. The complexity of implementing an OCP as described may be significantly greater than implementing a simple LTS. As you start to add optimisation variables or increase the complexity of the underlying models, this technique quickly turns into a highly complex and computationally-hungry problem.</p><p>Optimal control can be divided in several classifications such as problem classes, problem formats and solver classes. And within these there will be continuous and discrete problems, online and offline controllers, etc. But the general idea is that: <strong>Problem classes are problem concepts that can be represented with a mathematical or problem format and that problem can then be fed to a solver to get a result</strong>. I would recommend looking as CasADi since it is a widely used open-source tool for nonlinear optimisation [5].</p><ol><li><strong>Problem Classes (Conceptual Types)</strong></li></ol><ul><li>Static optimisation, optimal control problems (OCPs), graphic or route planning problems, task scheduling, stochastic/robust optimisation, multi-objective optimisation, game-theoretic or mutli-agent optimisation, etc.</li></ul><p><strong>2. Problem Formats (Mathematical Structure)</strong></p><ul><li>Linear programming (LP), quadratic programming (QP), nonlinear programming (NLP), mixed-integer linear and nonlinear programming, dynamic programming, stochastic programming, etc.</li></ul><p><strong>3. Solver Types (Algorithms / Tools)</strong></p><ul><li>local, gradient-based, IPOPT, KNITRO, BONMIN, heurstic or meta-heuristic solvers, etc.</li></ul><h4>System simulation</h4><p>The next type of simulation is called system simulation. And the main objective is not necessarily to do the fastest lap or minimize lap time, but rather to understand the behavior of a system. Engineers will usually model specific systems or subsystems of the car (such as the battery, the engine, the electric motor, the engine controller, brake by wire controller, etc.) and supply a set of control or inputs signals to drive the system.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BNI6DrKOTQm3W2f-GgCnrg.png" /><figcaption>Figure 4. Example of a electromechanical system implemented for Tec Racing when I was in university.</figcaption></figure><p>In the end, this may be a simulation of a single system (for example the inverter) or it may be a full vehicle with multiple systems simulated. These subsystems may be simulated in a multiphysics, multi body or analytical way, or even in a surrogate model fashion. Often times, engineers will even concatenate simulation tools such that the input to a system simulation (which becomes a transient simulation) is the result of a LTS (for example a speed profile).</p><p>I have previously mentioned surrogate model. And without going into much depth what this means is a representation of a model, but not the model itself. As an example, instead of simulating the performance of an inverter in your full vehicle model, you might want to have an independent simulation of the inverter in much detail and tiny time step, then produce a lookup table of how it behaves and use that table in your full vehicle simulation. This is a surrogate model.</p><p>And just as a bonus, system models in Simulink can use the Simscape library [6] to do a multiphysics simulation by incorporating electrical, mechanical and thermal domains. The multi body library further allows to import CAD models and simulate them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4Ow36J94KrkFVnhqaQFtGQ.png" /><figcaption>Figure 5. Vehicle multibody simulation in Simulink [7].</figcaption></figure><h4>CFD/FEA</h4><p>On to CFD and FEA simulations. The simulations I have talked about above are more related to operational engineering. Meaning more relevant towards performance, car setup, pre-event preparation, tyre science, strategy optimisation, etc. However, CFD and FEA is usually more relevant for design work. For example, aerodynamics design, powertrain thermal analysis, electronics thermal analysis and structural and materials analysis. But what do CFD and FEA actually mean?</p><ul><li>FEA – finite element analysis (used more commonly to evaluate loads and stresses on objects)</li><li>CFD – computational fluid dynamics (used to understand how fluids behave)</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yTTtmXLxmhezR5Xls-NUhQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0DM6fwTUoYxXr8TJWrAyKg.png" /><figcaption>Figure 6. CFD Example for Porsche 918 model in Star CCM+.</figcaption></figure><p>And how they work (in a very general simplification) is by taking a 3D model, breaking it down into tiny differentials, and solving the governing equations. To give you an example with FEA, a structural member of the chassis can be divided into small parts using a mesh and then governing equations (such as conservation of mass and energy, or relations between stress and loads) are evaluated in each element and iteratively, the impact of one element on its neighbors is evaluated to reach an acceptable solution (usually a consistent solution within a specified tolerance of error between iterations). As you can imagine, this simulation methodology is quite computationally expensive, often requiring multiple machines to work together, or at least multiple processors in a computer. CFD on the other hand will evaluate other governing equations that include Navier-Stokes equations to understand how a fluids (typically water, fuel or air) move around an object. This is very much what aerodynamicists use to evaluate their designs.</p><h4>Monte Carlo</h4><p>Next, a less known kind of simulation. Monte Carlo simulations essentially calculate the probability of an outcome. Basically, a way of predicting what is going to happen by modeling uncertainty. So, applied to motorsport, with a Monte Carlo simulation, you could simulate a specific race strategy and get the most probable position your will end up getting. How does it do that? Well, basically by trying out everything that could happen in a race. Essentially it does a large amount of iterations with random sampling and then creates the probability distribution for race results.</p><p>In race strategy Monte Carlo simulations, engineers can take into account the cars pace, tyre degradation, pit stop strategy, safety car probability, red flag probability and even weather to essentially run thousands of races with random combinations of those factors and then tell you what would be your most likely result as a function of statistics.</p><p>Another very attractive methodology is to use Monte Carlo simulations to actually give you the best race strategy. So, you could simulate thousands of race strategies, get a probability distributions for lap times and in that way choose the strategy that yielded the lowest lap times. For more robustness and realism, other techniques like game theory to incorporate the different players or drivers in the race might be needed.</p><h4>X-in-the-loop</h4><p>The final simulation type I will discuss in this post is X-in-the-loop. And the reason why it’s called <strong>X</strong>-in-the-loop is becuase it can be hardware-in-the-loop (HIL), software-in-the-loop (SIL), model-in-the-loop (MIL), driver-in-the-loop (DIL), etc. where the objective is to simulate a specific situation to put a device or model under test. And the device under test (DUT) will change to be a microprocessor, a human driver, a model, the embedded software of an ECU, a motor, etc.</p><p>In my professional experience, I have had to work on developing HIL environments from scratch and the possibilities that they open are inmense. From automating design validation tests, to testing new candidate designs, to exploring and reproducing faults to improve designs. And this can be applied not only to physical designs (PCBs, ECUs, etc.), but also to the embedded software within them — either by what we call unit tests (MIL) or SIL (where the software as a whole is tested).</p><p>DIL simulators specifically present a huge advantage in terms of motorsport engineering for teams, because it allows them to put a virtual car to test with a real driver in the loop with the objective of fine tunning car models, creating setups for races, or validating car concepts of the future. Alternatively, it also allows to use a specific car and track model and put the real driver to the test. This is how a lot of rookie drivers train for tracks where they have never driven before, or how teams with driver academies prepare their junior drivers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/702/0*IjUnTLvAXemVfw0g.jpg" /><figcaption>Figure 7. Driver-in-the-loop simulator example [8].</figcaption></figure><h3><strong>Let’s connect!</strong></h3><p>Engineering is for everyone. I hope this article was helpful and informational. My objective here is to help students, fans and other engineers get a closer technical view of motorsports. If you have experience with any of the simulation methods I have mentioned, think I missed something or would simply like to share your experience or thoughts, please drop me a comment or message and let’s connect!</p><h3>Conclusion</h3><p>So, as you can see, there are many types of simulation techniques that are used in motorsports and Formula 1. From design, to testing, to vehicle performance and strategy, simulation is deeply embedded in the engineering workflow and decision making process of F1 teams. There is a clear advantage of running these simulations given how expensive track data is and how robust and how accurate simulations can get. I think as we continue to move more and more into the digital world and as more powerful technologies emerge and develop, simulation will continue to strengthen how we make decisions and how we approach new problems. Data-driven decision making is what enables new designs to be better and better.</p><p>In this series of blogs about simulation in racing, I will continue to explain the simulation types I have briefly mentioned in this post and hopefully bring you closer to the engineering behind them. In the next blog post, I will discuss the role of AI and Machine Learning in state-of-the-art simulation technology and workflows. I have worked in research groups where we develop digital twin technology so I will talk about this as well.</p><h3>Disclaimer</h3><p>I am a developing engineering in the field so this is me sharing with you some of my personal experience over the years in academia/research, personal projects, and professional experience.</p><p><em>This article is unofficial and is not associated in any way with the Formula 1, or any related companies and/or teams. The comments expressed in this article and in the analysis are personal and do not represent the position of any company and/or team. This article is for fan use and educational use. The content and materials are intended solely for educational and informational purposes.</em></p><h3><strong>References</strong></h3><p>[1] <a href="https://www.dynisma.com/dmg-1">DMG-1 Motorsport Simulator | Dynisma Motion Generators</a></p><p>[2] <a href="https://www.bbc.co.uk/sport/formula1/articles/cly24p9le0lo">Audi to be granted increased budget cap on entering F1 from 2026 season — BBC Sport</a></p><p>[3] Eshof, E., Kampen, J. and Salazar M. (2025) <em>A Computationally Efficient Framework for Free-trajectory Minimum-lap-time Optimization of Racing Cars</em>. Optimization and Control. doi: 10.48550/arXiv.2511.13522</p><p>[4] Halkiopoulos, M. (2020) <em>OpenLAP-Lap-Time-Simulator</em>. MathWorks File Exchange. Available: <a href="https://uk.mathworks.com/matlabcentral/fileexchange/75063-openlap-lap-time-simulator">OpenLAP-Lap-Time-Simulator — File Exchange — MATLAB Central</a></p><p>[5] <a href="https://web.casadi.org/">CasADi</a></p><p>[6] <a href="https://uk.mathworks.com/products/simscape.html">Simscape — MATLAB</a></p><p>[7] <a href="https://uk.mathworks.com/solutions/physical-modeling/simscape-vehicle-templates.html">Vehicle Simulation — MATLAB and Simulink — MATLAB &amp; Simulink</a></p><p>[8] <a href="https://www.automotivetestingtechnologyinternational.com/industry-opinion/how-to-harness-driver-in-the-loop-simulators.html">How to harness driver-in-the-loop simulators | Automotive Testing Technology International</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a9af3fcd5ae0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My first blog post]]></title>
            <link>https://medium.com/@diegobosa/my-first-blog-post-1beab06ed8e6?source=rss-e703c3132b4a------2</link>
            <guid isPermaLink="false">https://medium.com/p/1beab06ed8e6</guid>
            <category><![CDATA[motorsport]]></category>
            <category><![CDATA[optimal-control]]></category>
            <category><![CDATA[formula-1]]></category>
            <category><![CDATA[engineering]]></category>
            <category><![CDATA[energy-management]]></category>
            <dc:creator><![CDATA[Diego Botin]]></dc:creator>
            <pubDate>Sat, 11 Jan 2025 18:38:07 GMT</pubDate>
            <atom:updated>2025-01-11T23:17:50.740Z</atom:updated>
            <content:encoded><![CDATA[<p>My name is Diego Botin and I am new to writing blogs but I made it part of my New Year’s resolutions to share my work, interests and experiences with the objective of helping other people and creating a community. I have always been of the idea that the best way to learn is through other people’s experiences so you don’t necesarily make the same mistakes and it makes the whole process more efficient. Of course, you have to fail and make mistakes to learn for yourself, but knowledge is about building from what is already out there and eventually creating something new.</p><p>Whether you are a motorsport or F1 fanatic, enthusiast, student or already have years of experience in the industry, I hope my experiences and the projects I will be sharing will be of use or interest to you. After all, it is of great interest to me so if you also have any questions, suggestions or recommendations, feel free to reach out!</p><h3>About myself</h3><p>I was born in Cali, Colombia and lived most of my life in different cities in Mexico (Mexico City, Queretaro and lately Monterrey). I studied the bachelours of Mechatronics Engineering in Mexico’s Tecnológico de Monterrey which is not very related to what I currently do, as the vast majority of my courses were focused around industrial process automation, industrial networks and robotics controls. I didn´t have many courses directly related to automotive engineering, let alone to motorsports. After all, Mexico has no motorsport engineering industry. So how did I manage to get into F1?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*hYU6xY_4y18_xRsr" /><figcaption>Tecnológico de Monterrey campus in Monterrey with a view of the Cerro de la Silla.</figcaption></figure><p>Well, I focused on gaining all the knowledge and practical experience that I could get outside university courses. So, I was part of the university’s scuderia (Tec Racing), and I also did some profesional experience working part time as a Jr. Research Engineer in the university. These two experiences brought me invaluable experience in race car design, race team management, engineering design and manufacturing, research methods, leadership and project planning. I also prepared myself for what I wanted to do A LOT. I would search job positions that I wanted to do in the future, and looked at the job description and requirements. This is how I knew I needed to learn and get good at Matlab and Simulink. Eventually I got an offer to intern in Red Bull Powertrains for a year in the UK and that was my start in F1. Let me tell you a little about each of the experiences.</p><h3>My Time In Tec Racing</h3><p>As soon as I started my bachelours course in Tec de Monterrey, I got involved with what would later become Tec Racing. The scuderia had 4teams where students designed and raced vehicles in different competitions such as Baja SAE, Formula SAE, Electraton and Gravity. I worked a little on the Gravity car which is a car with no motor, relying soley on gravity to race down a mountain (Yes it’s dangerous to race a student-designed car down a mountain at up to 100 km/h but that’s what makes it fun). Don’t worry, we always ensured the integrity of the car with good engineering practices and the safety of the driver. In Gravity, we won second place for two categories and first place for engineering design.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ktrV0I35PE-E-SAJX9B4xg.jpeg" /><figcaption>My last race with Electraton in Monterrey.</figcaption></figure><p>However, my focus first started with Electraton where I developed projects such as a designing a velocimeter, designing the steering wheel with driver controls, engineering the sensor network, and creating and improving a vehicle simulator for systems design validation. I carried out multiple roles and “went up the ladder” until I reached Chief Engineer of Tec Racing where I was coordinating the engineering aspect of the scuderia.</p><p>This was were I was able to learn and develop motorsport engineering skills so it was a big part in my uni experience. I also made the best friends throughout my time in uni. Definitely a win-win.</p><h3>Research Experience</h3><p>After my time in Tec Racing, I got a part-time job in one of my university’s research group. I founded a new project where we focused on developing digital twins for vehicles and for urban spaces. I was also doing project management for other projects in the group. During my time in research, I learned so many skills about research, management, project planning and team building. This is where I also discovered how much I like to research and share my findings.</p><p>I also got the chance to publish some conference and journal articles both as main author and co-author. Feel free to read some of my work <a href="https://www.researchgate.net/profile/Diego-Botin?ev=hdr_xprf&amp;_tp=eyJjb250ZXh0Ijp7ImZpcnN0UGFnZSI6ImhvbWUiLCJwYWdlIjoiaG9tZSIsInBvc2l0aW9uIjoiZ2xvYmFsSGVhZGVyIn19">here</a>.</p><h3>Placement Year</h3><p>After the experiences I had, I managed to get a placement year experience in none other than the Red Bull F1 team! I worked in electrical design in Red Bull Powertrains where we were developing the next generation Formula 1 2026 powertrain with incredibly talented engineers and placement students. During my time there, I worked about half of the time in actual PCB design and the other half in system simulation for the energy recovery system (ERS) and developed hardware-in-the-loop (HIL) testing models for some PCBs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*TXQlAJOfUZaC2PCRnB7-Vg@2x.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*kDs6X-AnDTYM6V_BhJOZlQ@2x.jpeg" /><figcaption>Some memories from my time at Red Bull.</figcaption></figure><p>It was an amazing experience where I learned a lot about what I was doing, but also learned so much about how F1 teams work, F1 itself and how to cope with stress and an intense working environment while trying to learn and deliver high quality work.</p><p>After this experience, I went back to uni to graduate and ended up getting my current job as Development Engineer in Haas F1 Team where I work and continue to learn every day about the car’s electronic control unit (ECU), vehicle control and HIL development. It truely is a team where I feel valued and excited to go to work every day. Plus we just had one of the most succesful seasons the team has seen!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mbiT-6mRtx8Ju582t5LmMg@2x.jpeg" /></figure><h3>Projects</h3><p>In my profile, I will be posting about projects I undertake from my masters program as well as some of the projects I like to work on in my free time. Don’t worry, I don’t plan on going too deep into the maths behind everything, but just enough to give a complete idea of how the projects are being developed and why. After all, the idea is that other people can take inspiration and guidance to start and develop their own projects. I would love to hear what you are working on as well! Some of the projects I will be working on are:</p><ol><li><strong>Optimal energy use</strong> — The objective of this project is to create a set of classes to solve the optimal control problem (OCP) of minimum lap time with an optimal use of energy for a race car. I will be basing the system in a Formula 1 car with 2026 regulations, but it should be flexible and robust enough to provide the same functionality for other vehicles to be simulated on different tracks. This project has already started and my next post will be about the first part.</li><li><strong>Model-based energy management controller simulator </strong>— The idea is to design a Formula 1 model in Simulink where a real-time energy management controller can be implemented and simulated. I have never worked on energy management but have read a lot about it so it should be interesting.</li><li><strong>Real-time vehicle simulator (pseudo-driver-in-the-loop)</strong> — I just discovered that I could control a vehicle model through an Xbox controller in Simulink. So the next logical question is: why not design a controller-in-the-loop simulator for a race car? I think this project will take a long time as I will not be investing too much time into it, but I may show some progress at some point.</li></ol><p>Let me know if you guys have questions about subjects I might be able to talk about!</p><h3><strong>Contact me</strong></h3><p>LinkedIn: <a href="https://www.linkedin.com/in/diego-botin/">https://www.linkedin.com/in/diego-botin/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1beab06ed8e6" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>