Enhancing Autonomous Racing Cars with a Compact and High-Performance Compute Unit
The rise of autonomous racing cars in motorsport highlights the potential of driverless technology in extreme environments. This field demands exceptional speed and control precision. Current compute units face challenges: high-performance units are bulky and affect speed, while compact units lack sufficient computational power for quick response and precise control. These issues hinder performance enhancement in autonomous racing cars. The advent of high-performance, compact embedded devices presents new opportunities. Australia’s QUT Motorsport team has adopted the powerful LattePanda Sigma single-board computer as their compute unit, significantly boosting their car’s speed.
Figure: Autonomous Racing Car Developed by QUT Motorsport
LattePanda Sigma is a highly integrated, powerful single-board computer, ideal for high-performance computing in space-constrained applications. It features the 13th generation Intel processor and up to 32GB RAM, offering outstanding computational and data processing capabilities. Its compact size and rich interfaces allow easy integration into complex systems, providing unmatched advantages in rapid response and high-precision control for autonomous racing cars. In the following sections, we will explore its applications in driverless racing cars through specific case studies.
LattePanda Sigma in the QEV-3D Autonomous Racing Car
QUT Motorsport and QEV-3D
QUT Motorsport is comprised of the most talented and dedicated students in Australia, who are passionate about producing innovative solutions for Electric and Autonomous Formula Student race cars. As the team said:“ We are committed to refining our expertise through design and fabrication and are determined to produce the best race cars each year.” Currently, their lineup comprises both human-driven and autonomous racing cars. The QEV-3D model, which is their extensively engineered autonomous vehicle, is equipped with the LattePanda Sigma as its compute unit.
Figure: QUT Motorsport from Australia
Figure: QEV-3D Autonomous Racing Car from QUT Motorsport
The electric-autonomous car, QEV-3D serves as a testbed for our autonomous racing system. Utilizing the chassis of their 2020-2021 Electric Competition car, QEV-3, they have retrofitted it with the mechanical and computational systems required for driverless operation. Their autonomous system utilizes computer vision, machine learning, and many other onboard sensors to accurately navigate a track.
The Role of LattePanda Sigma
When seeking the compute unit, the objective of the QUT Motorsport team is very explicit: "Our race car needs to be compact to ensure it’s light and aerodynamic, so we were looking for the smallest SBC that could perform the high compute load of autonomous racing." Later they found that the LattePanda Sigma might be a perfect fit for their racing car.
The LattePanda Sigma is used as the central brain for their autonomous race car. Its main role includes:
· Central Control Unit: It handles core data processing tasks. It integrates and coordinates various subsystems, ensuring seamless operation and communication.
· Data Processing: It processes sensor data from LiDAR (Light Detection and Ranging) and GPS to map the race track and navigate through it.
· Task Control Interface: The team interfaces with and controls other vehicle electronics using USB-CAN through the LattePanda Sigma to perform various tasks.
· Telemetry Transmission: The LattePanda Sigma transmits telemetry data over a local wireless network, allowing the team to monitor faults and view live sensor data trackside.
Why LattePanda
1. Compact Design
The size and weight of the compute unit are critical in the design of racing cars, where every gram counts towards achieving optimal speed and aerodynamics. Previously, QUT Motorsport used a high-performance industrial-grade robotic controller (ADLINK ROScube), which, while powerful, was considerably larger and heavier. The LattePanda Sigma offers a significant reduction in size (146mm*102mm) and weight (305 grams) compared to their previous computing solutions.
This compact form factor allows it to be seamlessly integrated into the limited space available in the racing car, ensuring that the vehicle remains lightweight and maintains its aerodynamic efficiency. By minimizing the space occupied by the compute unit, the team can allocate more room for other essential components or reduce the overall weight of the vehicle. This reduction in weight directly translates to improved acceleration and handling, ultimately enhancing the car's speed and performance on the track.
In addition, the smaller size simplifies the design and assembly process, making it easier to fit the compute unit into the car's chassis without extensive modifications. This ease of integration not only saves time but also reduces the complexity of the overall design, allowing the team to focus on other critical aspects of the car's performance.
Figure: The Previous Compute Unit (Under the LattePanda Sigma)
Figure: LattePanda Sigma inside the QEV-3D Racing Car
2. High Performance
Autonomous racing requires real-time processing of vast amounts of data from various sensors, including LiDAR and GPS. The LattePanda Sigma, equipped with the 13th Intel processor (12-core, 16-thread) and up to 32GB RAM, provides the necessary computational power to handle these high-speed optimizations. Its ability to process sensor data quickly and accurately is crucial for navigating the race track and making split-second decisions, which directly impacts the car's performance and competitiveness.
While there are other small SBCs available, the QUT Motorsport team often falls short in terms of processing power and the ability to run x86 code natively. As QUT Motorsport noted, "What we run requires quite a bit of processing power and the ability, so many SBCs were off the table." The team found that realistically, no single embedded computer could run everything they needed without bottlenecks. Processing power is a critical factor in choosing the LattePanda Sigma.
Figure: LattePanda Sigma Features the 13th Intel i5-1340P Processor
3. Strong Expandability
The LattePanda Sigma boasts a wide range of interfaces and connectivity options, such as USB-CAN, enabling QUT Motorsport to seamlessly integrate and communicate with various auxiliary control electronics and sensors. This expandability ensures that the team can easily add or upgrade components as needed, enhancing the car's functionality while optimizing performance and safety in the high-pressure environment of racing. Through USB-CAN, the LattePanda Sigma can connect to a variety of peripherals, including LiDAR sensors for obstacle detection, GPS modules for precise positioning, IMUs for monitoring orientation and movement, and cameras for visual data processing. This integrated system allows QUT Motorsport to seamlessly incorporate new technologies and make rapid, strategic upgrades, continuously enhancing the car's performance.
Figure: Rich Interfaces of LattePanda Sigma
4. Multi-System Support and Software Compatibility
Running on Ubuntu 22.04.4 LTS with Docker and ROS2, the LattePanda Sigma as an x86 single board computer supports the autonomous software stack required for the car's operations. This compatibility with widely-used software platforms ensures that the team can leverage existing tools and frameworks, including YOLOv8 for real-time object detection, for development, testing, and deployment. This integration streamlines their workflow and reduces the time needed to bring new features to the car.
Figure: QEV-3D Autonomous Racing Car in Testing
Conclusion
The implementation of the LattePanda Sigma has significantly elevated the capabilities of the QEV-3D Racing Car project, proving to be a robust solution that seamlessly balances size and performance. The QUT Motorsport team indicated the QEV-3D is currently being readied for competition under the F-SAE autonomous racing category, alongside THE human-driven car QEV-4, to compete in the Formula Student Australasia Competition in December, 2024. We hope that the LattePanda Sigma will aid them in achieving outstanding results. If your project also demands a compact yet powerful computing solution, the LattePanda Sigma might be your next strategic move.
