Executive Summary
The Indy Autonomous Challenge (IAC) AI Racing Tech (ART) team uses the AUTERA AutoBox to log sensor data from the Dallara AV-21 autonomous race car. These logs are then ingested into Amazon Simple Storage Service (Amazon S3) in the Amazon Web Services (AWS) cloud. These S3 buckets can be used in the AWS computing service Elastic Compute Cloud (EC2) to improve the team’s autonomy stack performance. The optimized code is redeployed on the AUTERA for racing success.
Challenges
Being distributed in this manner, the ability to collaborate to develop and test in a cloud environment while having confidence that the solution will transfer to the in-vehicle PC is essential. The team does not have unlimited access to the actual racecars or a single lab for on-premises installations.
Traditional alternatives such as duplicating racecars or setting up multiple on-premises installations are not practical due to cost and budget constraints at the university and team level. This Sim to Real solution is necessary for a team without unlimited access to a real racecar on track and without a single lab for on-premises installations of a virtual architecture.
Partner Solution
The data collection process begins on the track, where the ART team captures valuable information in the form of rosbags using the removable 16 TB SSD available on the AUTERA AutoBox. These rosbags are then ingested into AWS S3 buckets, which serve as secure storage for the team's data. Leveraging the power of Amazon Web Services (AWS), the team can utilize the S3 buckets within EC2 instances to run multiagent simulations, allowing them to fine-tune their autonomy stack.
By employing multiagent simulations in the cloud, the ART team can efficiently iterate and optimize their code. They can simulate various scenarios, test different algorithms, and analyze the performance of their autonomous driving system. This cloud-based approach enables the team to refine their solution without requiring constant access to the physical racecars, making the most of their limited track time.
Once the code has been optimized and validated through simulations, it is deployed into a similar environment on the AUTERA AutoBox, which is securely installed within the cockpit of the AV-21 racecars. This process ensures that the developed solution translates effectively from the virtual simulations to the real-world racecar, further enhancing the team's confidence in the performance of their autonomous driving stack.
Results and Benefits
The #1 ranked US team in the 2022-2023 season finished 2nd at the Indy Autonomous Challenge presented by Cisco at Texas Motor Speedway. This was the first race using AUTERA AutoBox. Cloud-based simulations offer dispersed developers the opportunity to collaborate effectively.
Indy Autonomous Challenge
The Indy Autonomous Challenge (IAC) AI Racing Tech (ART) team is developing an autonomous driving stack for successfully navigating their AV-21 Dallara Indy Light racecar. The teams only have limited track time with the AV-21s. Additionally, the team is scattered across the continental United States and beyond, with member universities ranging from Carnegie Mellon in Pittsburgh, PA to University of California schools in San Diego and Berkeley, and all the way to the University of Hawaii.