In 2022, the Indy Autonomous Challenge (IAC) hosted a groundbreaking event at the Las Vegas Motor Speedway. Two autonomous racecars competed against each other during the Consumer Electronics Show (CES). At CES 2025, the (IAC) and its member teams made history again. For the first time ever, four autonomous race-cars braved the chilly weather and high winds to compete and race against each other. Robotic competitions seem, well, robotic and unexciting – but IAC competitions invoke the same passion and emotion as human race-car competitions like the Indy 500, NASCAR and Formula 1.
Location, Teams, Racecars, AI Driver
The event was held on 9th January, 2025 at the Las Vegas Motor Speedway, located ~ 20 miles from the Las Vegas Convention Center (LVCC) that hosted the annual Consumer Electronics Show (CES). A significant portion of the CES exhibition and talks focused on robotics, artificial intelligence (AI) and autonomous machines (cars, trucks, mining and agricultural vehicles) and the sensors and compute systems that power them. IAC hosted a couple of round-table discussions at their booth in the LVCC that focused on physical AI, and promoted the autonomous racecar event at the Las Vegas Speedway, an asphalt surface, 1.5 miles long with 20-degree banked turns and 9-degree banked front and backstretch. It seats 123,000 people.
IAC event competitors include university teams from the USA, Germany, Italy and Korea. A total of 9 teams participated in the 2025 Las Vegas competition:
- Cavalier Autonomous Racing (University of Virginia)
- PoliMOVE-MSU (Collaboration between Politecnico di Milano, Italy, Michigan State University and University of Alabama)
- KAIST (Korea Advanced Institute of Science and Technology)
- Unimore Racing (University of Modena and Reggio Emilia, Italy)
- Autonomous Tiger Racing (Auburn University)
- Purdue AI Racing (Purdue University)
- AI Racing Tech (Universities of California at Berkeley and San Diego, Hawaii and Carnegie Mellon)
- IU LUDDY (Indiana University)
- Caltech
The cars used by the teams are the IAC AV-24, with a Dallara built chassis used in the IndyNXT Championship Series. This is modified by IAC to install the robotics hardware, sensors and compute stack. All cars identical in hardware, the performance differentiation is driven by the the AI driving stack that each team develops and installs on their cars. In competition, the AI driver is completely autonomous with no real time communications with humans (other than safety and race personnel). The quality, performance and speed (latency) of the AI stack is directly dependent on how well each of the following sub-stacks perform:
- Localization: knowing the car position relative to the track, critical for making decisions on negotiating turns and maximizing speed on straight sections. GPS sensors, cameras and LiDAR provide this capability.
- Perception: situational awareness of the vehicle surroundings including obstacles and other racecars is especially important in multi-car races. The ability to discriminate relevant sensor data (LiDAR, radar, camera) and process it rapidly allows for low control latency.
- Physical Modeling of Vehicle Dynamics: Physical models of vehicle dynamics under different conditions of track conditions and configurations, weather, wind, ambient and tire temperature are used along with sensor data (pressure and temperature sensors) to adjust algorithms for vehicle control and push the limit in terms of “safe” speed. Refining these models in terms of fidelity and accuracy is part of the Sim-to-Real revolution in physical AI implementations.
- Path Planning & Vehicle Control: items 1-3 provide inputs for path planning and control of the vehicle and decisions on speed, direction, steering and braking.
- Strategy and Tactics: Each team develops its own strategy for winning by using simulations, physical trials on the track, understanding the strength and weaknesses of other teams and building this strategy into the AI driver.
The 4-car Passing Event
The 4-car event (IAC prefers to call it a exhibition race rather than a competition!), the first ever with high speed racecars on a closed track was thrilling and displayed the above 5 capabilities of the AI Driver to varying degrees. The competing teams were chosen based on their prior performance and experience:
- Cavalier Autonomous Racing
- PoliMOVE-MSU
- KAIST
- Unimore Racing
The focus of this race was to test AI drivers at high-speed in a multi-agent environment where perception and race strategy are especially important. An important consideration is that none of the 4 cars know exactly their competitive positions and ranking during the race (especially during the later laps). Effectively, the AI Driver is programmed to maximize performance with situational constraints (obstacles, neighboring cars locations, weather, wind, etc.). Two important features/rules were instituted:
1) Due to the weather, cars were restricted to run at a maximum speed of 80 mph unless they were passing other cars
2) Each car was allocated a total time budget for a feature called “Push to Pass (P2P)”. This is a well known feature in professional human racecar events like the Indy 500 and NASCAR where propulsion and aerodynamic drag control can be activated to provide acceleration bursts at certain points in time in order to overtake opponents. When activated, P2P enables the IAC cars to hit peak speeds of 120 mph.
Dr. Madhur Behl, the faculty advisor for the Cavalier Racing Team explained the specific P2P rules for the 4-car racing event:
1) Cars are allowed a 2-second burst of 100% throttle, followed by 28 seconds of boost throttle (35% throttle), with a 20-second cooldown period between uses. Effectively, a P2P activation could range from 22-50 seconds (the shorter activation uses only the 2 second burst followed by 20 seconds of cooling)
2) Each vehicle was allocated a time budget to use the P2P feature depending on its starting position (inside vs outside track and location on track, with sufficient gaps for safety). PoliMOVE and Unimore were on inside track with P2P budgets of 300 s and 420 s respectively. KAIST and Cavalier Racing were on the outside tracks and had budgets of 360 s and 480 s respectively. These positions and P2P time budgets were known to the teams well in advance of the race and were built into the race strategy.
With a track length of 1.5 miles, and maximum speeds of 80 mph, 20 laps are expected to take ~30 minutes (earlier laps are slower to allows tire warm-up and traction), with later laps taking about a minute each, allowing the teams to decide how to use their allocation of P2P time (and ensure cooling time between P2P activations).
Of the 4 teams, PoliMOVE unfortunately experienced a mechanical/communication failure during the first 2 laps, which they fixed quickly and joined in lap 3. Although they could not really win, they participated in the rest of the race, keeping intact the multi-agent configuration (the other 3 cars had no way of knowing that this happened). KAIST led from laps 3-14 by utilizing their P2P budget fully (maybe reasoning that by doing this, other cars would not be able to catch up due to weather or mechanical reasons ?), but this effectively meant that they had no P2P available in the last 6 laps. At this point, Cavalier Racing overtook KAIST, and shortly thereafter, so did Unimore. Both team strategies were to use P2P at the right times and location, and varying the amount of throttle. After this, it was a battle between the two, with Unimore eventually pulling ahead at the finish line by 2 car lengths (Figures 1 and 2):
Individual Speed Trials and 2-Car Passing Competition
The 4-car passing exhibition event was preceded by 2 other events:
1) Individual Speed Competition in which single teams at a time participated. The teams competing here included Autonomous Tiger Racing (Auburn University) which reached the winning average lap speed of 164 mph, followed by IU Luddy (Indiana University) at 158 mph and Caltech clocking in at 144 mph.
2) 2-Car Passing Competition between Purdue AI Racing (Purdue University) and AI Racing Tech (Universities of California at Berkeley and San Diego, Hawaii and Carnegie Mellon). AI Racing Tech won by overtaking AI Racing by passing at 162 mph in the final moments of the last passing round.
Aidoptation – Technology Transfer to a For Profit Entity
IAC announced an important development at CES – the formation of a for-profit entity named Aidoptation, headquartered in Belgium and focused on using the learnings from the non-profit IAC entity and its university team members to improve ADAS and autonomous driving at high speeds. In essence, the idea is to complement the commercial sector working on commercializing SDV (Software Defined Vehicles) and transportation autonomy, and provide the capability for handling edge cases encountered at high speeds – with best-in-class AI sub-stacks like vehicle dynamics, sensors, perception, localization and control.
One of the commercial products under consideration is the OpenDriver AI architecture which would utilize the best of the different AI sub-stacks from different university team members. IAC has essentially accumulated ~25,000 miles of high speed racing data which has been used by 100s of Ph.D. level researchers at these different university teams (each team has about 10 Ph.D. students). The intellectual property assets and experience in the area of high speed autonomy is unique and at this point, ripe for commercial deployment.
As Aidoptation progresses in its commercialization journey, the non-profit entity (IAC) will continue its charter of doing cutting edge research in the high speed autonomy arena, and provide test and evaluation services for its existing contract with DARPA on the TIAMAT (Transfer from Imprecise and Abstract Models to Autonomous Technologies) program.
Paul Mitchell who has shepherded the non-profit IAC entity will also serve as the CEO of the for-profit entity. “Autonomous high-speed physical AI has the potential to revolutionize mobility. With countless world records and multiple patents, IAC has led this field. We are launching Aidoptation as a commercial spin-out to build on these achievements and deliver real commercial solutions for managing high-speed edge cases that will unlock the full potential of autonomous mobility to speed up commerce and save lives.” The financial investors in this new entity include Belgium’s Sovereign Wealth Fund and Ethias, a leading insurance provider. The rational for locating in Belgium are to support Europe university teams as well as incubate talent and resources for progressing physical AI in Europe.
Going forward, apart from commercializing current and future IAC intellectual property, Aidoptation will also market, organize and monetize future racing events and take over the operational aspects of building, upgrading and maintaining robotic racecars for its university partners. The company is expected to start operations in early 2025.
Upcoming Events
Two upcoming events organized by IAC-AIdoptation are on the horizon:
1) On the heels of a record breaking speed event in 2022 at the historic Kennedy Space Center in Florida, the IAC will attempt to break the speed record set by PoliMOVE of 192 mph, the fastest achieved by a robotic race car to date. It will again occur at the Kennedy Space Center in concert with the 1000 Miglia Experience USA Florida 2025 event, between February 22-24, 2025.
2) The next IAC racing event will occur at the Monza F1 Circuit during the Milan Monza Motor Show (at the Autodromo Nazionale Monza racetrack) on June 27-29, 2025.
Autonomous racecar competitions have the potential to create best-in-class AI drivers, sensors and compute stacks that can be used to complement commercial ADAS and autonomy efforts in high-speed, on-road environments. Leveraging the learnings of 25,000 miles of high speed autonomy and the intellectual horsepower of 100s of Ph.D. level researchers over the past 5 years can create valuable opportunities for commercial ADAS and autonomous vehicle companies to understand edge cases and adapt these learnings to maximize safety and efficiency at ever increasing roadway speeds for autonomous cars.
