In August 2024, the California Institute of Technology (Caltech) announced its participation in the Indy Autonomous Challenge (IAC), the world's fastest autonomous racing program in partnership with Beyond Limits. Caltech's CAST team, known as the Caltech Autonomous Systems and Technologies (CAST) Racer, is led by esteemed faculty members, including Dr. Soon-Jo Chung, Dr. Fred Y. Hadaegh, and Dr. Mory Gharib. This effort was enhanced through joint development with Beyond Limits and the use of our Hybrid AI capabilities. Together, we aim to push the limits of autonomous racing and apply the learnings in real world environments. In this interview, our Beyond Limits CEO AJ Abdallat discusses the company's involvement in the Indy Autonomous Car Challenge, the unique challenges faced, and the broader implications for AI development.
Our goal is to push the boundaries of AI in real-world scenarios. Autonomous racing offers a unique testbed for extreme-edge AI systems. The challenge gives our team a chance to develop and refine control systems that can operate unmanned in high-speed, unpredictable environments. We believe this directly translates into industrial autonomous operation solutions that will deliver break-through efficiency and safety gains.
We also saw this as an opportunity to collaborate with top research institutions and industry experts. Autonomous driving is more than just cars—it’s about advancing AI to handle complex decision-making without human intervention. That is the future we’re building.
Beyond Limits has always been at the forefront of AI development, focusing on hybrid AI solutions that combine data-driven machine learning with knowledge-based reasoning. This race presents a perfect environment to stress-test these technologies under extreme conditions, ensuring their reliability in industrial applications. The ability to process real-time sensor data, make split-second decisions, and operate with minimal computational power aligns directly with the requirements of mission-critical industries such as energy and aerospace.
Additionally, autonomous racing allows us to enhance our AI’s adaptability. Unlike structured industrial settings, the racing environment is highly dynamic. Competing vehicles make unpredictable maneuvers, requiring advanced perception and real-time decision-making capabilities. By tackling these challenges, we refine our AI’s ability to function in complex, every changing settings, something that can be directly applied to optimizing industrial workflows, reducing downtime, and enhancing safety protocols.
Another key factor in our decision was the opportunity to test closed loop automation at scale. In many industrial sectors, AI still requires frequent human oversight and decision support. The autonomous car challenge demands complete reliance on AI-driven decision-making, reinforcing our mission to develop autonomous solutions that require little to no human intervention. This translates into practical applications such as remote facility management, closed loop process control, and autonomous industrial inspection and maintenance.
Participating in this challenge is also about talent development. Our collaboration with Caltech and other leading AI research institutions allows us to work with some of the brightest minds in autonomous systems. This knowledge exchange accelerates innovation, ensuring that Beyond Limits remains ahead of the curve in AI-driven automation.
Ultimately, this race is not just a competition, it’s an investment in the future of autonomous control. The lessons we learn here will shape AI’s role across multiple industries, helping businesses unlock new levels of efficiency, safety, and reliability.
Most autonomous vehicles operate under controlled conditions, well-mapped roads, predictable traffic, and access to cloud computing. Autonomous racing is a completely different challenge.
• Cars travel at speeds exceeding 150 mph.
• Decisions must be made in milliseconds with no external guidance.
• Systems must operate with minimal computational power while processing vast amounts of sensor data.
• The margin for error is nearly zero—one mistake can mean a crash.
Each of these factors raises the complexity of the task. Unlike street vehicles that rely on predefined routes, race cars operate in a fluid environment where speed and agility determine success. Sensors must interpret vast amounts of data in real time while the AI continuously recalibrates strategies. External input is unavailable, meaning the system must be fully self-reliant. The goal is to refine the AI’s ability to anticipate and react with precision, pushing it beyond conventional applications.
To add to the difficulty, environmental factors play a significant role. Small variations in track temperature, tire wear, and aerodynamics impact vehicle control. The AI must continuously adapt to these shifting conditions while maintaining peak performance. The ability to process incomplete or imperfect data, make real-time corrections, and optimize movement with limited computational resources is what sets this challenge apart.
Moreover, unlike traditional autonomous vehicles that rely on extensive mapping and high-resolution data from connected systems, these race cars must navigate based solely on onboard sensors. There is no cloud connectivity or external reference points, the AI must handle everything internally and react in a fraction of a second. This extreme environment forces AI to evolve, improving its ability to function under unpredictable and fast-moving conditions.
Competition not only pushes AI development but also highlights the importance of merging perception, control, and decision-making into a single, cohesive system. Each decision is crucial, misjudging a turn by a fraction of a second can result in failure.
We specialize in AI for critical industries—energy, manufacturing, and logistics. These sectors require autonomous systems that function in complex, high-stakes environments.
Take oil and gas, for example. Many operations occur in remote, hazardous locations. Automating control systems in these settings require AI capable of making decisions without human oversight. The technology we develop for autonomous racing can help:
• Improve AI-driven process control in industrial systems.
• Reduce the need for human intervention in dangerous environments.
• Enhance safety by enabling machines to detect and respond to risks in real time.
The ability to operate without human intervention is a common requirement across industries. Whether managing an offshore drilling platform or directing supply chains in a manufacturing facility, AI must assess risk, adjust for variables, and act decisively. By refining control mechanisms in the racing environment, we strengthen AI’s capacity to handle unpredictable conditions. The racetrack forces AI to solve problems instantly—a valuable skill when applied to sectors where efficiency and safety are top priorities.
Time was our biggest challenge. We had just a few months to assemble the car, integrate the hardware, and fine-tune the AI. Other teams had years of experience, but we had to start from scratch.
Some key hurdles:
• Sensor integration: Over 100 sensors had to be connected, calibrated, and optimized.
• Computational limits: The AI needed to make decisions instantly without high-powered cloud computing.
• Simulation vs. reality: We relied on simulation for training, but real-world track conditions often varied.
• Limited track time: Each team got only a few minutes of actual race testing. Every session had to count.
Working within strict timelines requires precision at every stage. The car’s sensors provided raw data but converting that into actionable intelligence required a methodical approach. Processing constraints meant selecting only the most relevant data points, discarding noise, and running calculations in fractions of a second. Since track time was limited, simulations had to be as accurate as possible, mimicking real-world conditions to refine performance. Every decision, from hardware selection to software adjustments, had to be deliberate to maximize our efforts.
Despite these obstacles, our car performed beyond expectations. In our first competition, we reached a top speed of 155 mph and secured a strong finish in our category.
The next milestone is the Monza race in Italy. Unlike the oval tracks we’ve competed on, Monza features tight corners and complex maneuvers. This will push our AI to new levels of precision and adaptability.
• Perception and localization: Enhancing how the car detects and responds to other vehicles.
• Real-time decision-making: Refining algorithms to handle rapid overtaking and obstacle avoidance.
• Simulation accuracy: Bridging the gap between virtual training and real-world execution.
Beyond this race, we’ll apply these learnings to industrial automation. The core technology—fast, autonomous decision-making—has broad applications beyond motorsports.
This project is about more than just racing. It’s about proving what AI can achieve in high-risk, high-speed environments.
By tackling one of the toughest challenges in autonomous technology, we’re gaining insights that will shape the future of AI across industries. Whether it’s redefining autonomous control in manufacturing, optimizing industrial operations, or enhancing safety in remote locations, the impact of this work extends far beyond the track.
We’re excited about what’s ahead and proud of what our team has accomplished in such a short time. The Indy Autonomous Car Challenge is just the beginning.
Follow our CEO Aj Abdallat on LinkedIn HERE
Want to know more about Beyond Limits - Contact us today
