Minimalistic Autonomous Stack for High-Speed Time-Trial Racing
Mahmoud Ali, Hassan Jardali, Youwei Yu, Durgakant Pushp, Lantao Liu · 2025
Read Minimalistic Autonomous Stack for High-Speed Time-Trial Racing by Mahmoud Ali, Hassan Jardali, Youwei Yu, Durgakant Pushp, Lantao Liu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-19636.txt.
Introduction
Self-driving race cars can top two hundred kilometers per hour, but testing them eats track time that few teams have. Mahmoud Ali and four co-authors present a slim racing stack built for fast deployment. Their system ran real speedway time trials after only eleven hours of track practice. Total track distance in testing stayed near three hundred twenty-five kilometers.
Peak speed reached two hundred six kilometers per hour. The team reports tracking accuracy, vehicle behavior, and safety margins in detail. A minimal design cut the usual long development cycle sharply. Their lessons target teams that must validate on real cars with scarce access. The work shows that careful integration can matter as much as algorithmic novelty in racing.
Catalog record
| CREATOR | Mahmoud Ali, Hassan Jardali, Youwei Yu, Durgakant Pushp, Lantao Liu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-19636
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Minimalistic Autonomous Stack for High-Speed Time-Trial Racing Mahmoud Ali, Hassan Jardali, Youwei Yu, Durgakant Pushp, Lantao Liu Abstract Autonomous racing has seen significant advancements, driven by competitions such as the Indy Autonomous Challenge (IAC) and the Abu Dhabi Autonomous Racing League (A2RL). However, developing an autonomous racing stack for a full-scale car is often constrained by limited access to dedicated test tracks, restricting opportunities for real-world validation. While previous work typically requires extended development cycles and significant track time, this paper introduces a minimalistic autonomous racing stack for high-speed time-trial racing that emphasizes rapid deployment and efficient system integration with minimal on-track testing. The proposed stack was validated on real speedways, achieving a top speed of 206 km h − 1 206\text{\,}\mathrm{km}\text{\,}{\mathrm{h}}^{-1} within just 11 hours’ practice run on the track with 325 km 325\text{\,}\mathrm{km} in total. Additionally, we present the system performance analysis, including tracking accuracy, vehicle dynamics, and safety considerations, offering insights for teams seeking to rapidly develop and deploy an autonomous racing stack with limited track access. I Introduction Autonomous racing has gained significant traction in recent years, advancing both research and real-world deployment in high-speed autonomy. Competitions such as IAC [ 1 ] and A2RL [ 2 ] provide a platform for testing cutting-edge autonomous systems in extreme conditions. These events have driven advancements in perception, planning, and control algorithms [ 3 , 4 , 5 , 6 ] , leading to fully autonomous race cars competing at 290\text{\,}\mathrm{km}\text{\,}{\mathrm{h}}^{-1} [ 7 ] . Despite these achievements, developing an autonomous racing stack for full-scale vehicles remains a resource-intensive endeavor due to the limited availability of dedicated racetracks for testing. Prior work has demonstrated impressive results but often relies on years of development and extensive track-testing time, making rapid deployment difficult for new teams. To address this challenge, we present a minimalistic autonomous racing stack designed for high-speed time-trial racing with a focus on single-car speed performance and rapid deployment. Our approach strategically maximizes track time utilization, enabling a fully functional autonomy stack with minimal on-track testing. The proposed system was implemented on the IAC AV-24 race car [ 8 ] and validated on real speedways, achieving a top speed of 206\text{\,}\mathrm{km}\text{\,}{\mathrm{h}}^{-1} with only 11 track hours and 325\text{\,}\mathrm{km} of practice runs. The key contributions of this work include: I) A minimalistic autonomous racing stack tailored for high-speed solo racing. II) A system integration of the proposed stack into the AV- 24 racing car. III) An evaluation of the proposed stack with emphasis on controllers performance and dynamics analysis. IV) A discussion of the safety measures incorporated into the system, along with an analysis of failure cases. Fig. 1 : IAC AV24 race car features a sensor-rich system to attain high-speed autonomy. II System In this section, we introduce the IAC AV-24 racing car, its sensor suite, and the actuation system, which serves as the primary interface to the car. II-A Hardware Overview The IAC AV24 autonomous racing car is based on the Dallara IL-15 employed in the Indy Lights Racing Series. The AV24 features a 2.0\text{\,}\mathrm{L} single-turbocharged, inline 4 -cylinder engine with a maximum power of 488 HP, paired with a six-speed sequential gearbox. The car is converted to be autonomous by replacing the cockpit with autonomous driving components, including a computing platform, sensors, an actuation system, and communication apparatuses. The vehicle’s actuation is enabled through a Schaeffler Paravan “SpaceDrive II” drive-by-wire (DBW) system. The AV24 is equipped with a dSpace AUTERA AutoBox PC, featuring an Intel Xeon CPU, an Nvidia A5000 GPU, and 14 TB of storage capacity. The AV24 is also equipped with a localization suite comprising three Global Navigation Satellite Systems (GNSS): two Novatel PwrPak7D (dubbed Novatel-A and Novatel-B) and one VectorNav-310. The AV24 perception stack includes six cameras—two front-facing, long-range cameras configured as a stereo setup and four wide Field-of-View (FoV) cameras providing 360 ∘ coverage—along with two Continental ARS540 4D radars (one front and one rear), each featuring a 60 ∘ FoV, as well as three Luminar Iris LiDARs, each with a 120 ∘ horizontal FoV to ensure complete 360 ∘ coverage. Fig. 1 shows our race car with various sensors with real data from exteroceptive sensors, including the front Radar, left LiDAR, and rear camera. Fig. 2 : System diagram. II-B Actuation Interface Systems The actuation system at a lower level comprises an embedded platform that includes Electronic Control Units (ECUs), servomotors for steering, throttle, and braking, a voltage management system, and a suite of monitoring sensors measuring values such as engine oil temperature and pressure, coolant temperature, tire temperature and pressure, engine RPM, wheel speed, wheel torque, and battery voltage. The New Eagle GCM 196 Raptor is used as the interface to this lower-level system, simultaneously handling the transmission of throttle, brake, gear-shift, and steering commands to the DBW system and routing sensor data to the autonomy stack. This interfacing is accomplished through the Controller Area Network (CAN) serial bus system. The lower-level software implements a state machine that governs the vehicle’s operational status throughout its run. This state machine transitions between states by verifying key functionalities, such as performing an actuation test, to ensure proper steering and braking. Additionally, it continuously monitors communication between the Raptor and the onboard computer via a rolling counter mechanism. Once all functionality tests have been passed, the engine can be started, allowing the vehicle to move into higher-level states (e.g., driving mode) and begin receiving commands from the autonomy stack. In the event of anomalies (e.g., a lost connection between the Raptor and the autonomy stack), the system initiates a supervised stop or an emergency shutdown. Another component interfacing with the lower-level system is Race Control, serving as a supervisory entity responsible for managing the race by issuing command flags that instruct the vehicle to stop, adjust speed, or return to the pit. Two categories of flags are employed: vehicle flags, which direct an individual car to start its engine, enter the pit lane, or come to a complete stop; and track flags, which are broadcast to all cars and, e.g., define the maximum allowable speed. These flags are transmitted to AV24 via a Marelli-powered system. III Minimalistic Autonomy Stack We describe our approach to implementing a C++/ROS2-based, minimalistic, efficient, and safety-critical autonomy stack for high-speed time-trial (single-car) racing. We also discuss the challenges encountered, including sensor driver integration, adherence to execution deadlines in estimation, planning, and control modules, as well as data recording, visualization, and communication between the stack and both the lower-level system and the base station. III-A Autonomy Stack Peripherals III-A 1 Raptor to Autonomy Stack Communication A DBC (Database CAN) file provided by the competition organizers is used to interpret the CAN messages. To integrate these messages within the stack, we leverage the raptor can driver [ 9 ] , to convert the CAN messages coming from raptor into ROS messages and publish them on multiple topics, as well as convert control commads from ROS messages to CAN messages to send it to raptor. III-A 2 Autonomy Stack to Base Station Communication We use Secure Shell (SSH) to remote access the onboard computer to launch our stack. However, rather than relying on the ROS2 framework for data exchange between