Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

IIT Bombay Racing Driverless: Autonomous Driving Stack for Formula Student AI

Yash Rampuria, Deep Boliya, Shreyash Gupta, Gopalan Iyengar, Ayush Rohilla, Mohak Vyas, Chaitanya Langde, Mehul Vijay Chanda, Ronak Gautam Matai, Kothapalli Namitha, Ajinkya Pawar, Bhaskar Biswas, Nakul Agarwal, Rajit Khandelwal, Rohan Kumar, Shubham Agarwal, Vishwam Patel, Abhimanyu Singh Rathore, Amna Rahman, Ayush Mishra, Yash Tangri

TL;DR

The paper presents IIT Bombay Racing's Driverless stack for Formula Student AI, addressing autonomous racing with a sensor-rich architecture that fuses LiDAR and stereo/monocular vision for cone perception, EKF-SLAM for localization, and Delaunay-based path planning with Stanley and Pure Pursuit control. A three-tier perception pipeline integrates LiDAR depth, monocular cues, and stereo depth to handle diverse scenarios, while EKF-SLAM (with MRPT) provides robust localization and mapping, and the Delaunay-based planner yields a racing line with reduced curvature. The system is implemented in ROS2/CAN, validated on a scaled robot and FSDS/EUFS simulators, and demonstrates reliable performance across perception, localization, and control tasks, including edge-case handling like occlusions and fallen cones. The work offers a practical, reliable autonomous racing stack for Formula Student AI events with clear pathways for further robustness in real-world edge cases.

Abstract

This work presents the design and development of IIT Bombay Racing's Formula Student style autonomous racecar algorithm capable of running at the racing events of Formula Student-AI, held in the UK. The car employs a cutting-edge sensor suite of the compute unit NVIDIA Jetson Orin AGX, 2 ZED2i stereo cameras, 1 Velodyne Puck VLP16 LiDAR and SBG Systems Ellipse N GNSS/INS IMU. It features deep learning algorithms and control systems to navigate complex tracks and execute maneuvers without any human intervention. The design process involved extensive simulations and testing to optimize the vehicle's performance and ensure its safety. The algorithms have been tested on a small scale, in-house manufactured 4-wheeled robot and on simulation software. The results obtained for testing various algorithms in perception, simultaneous localization and mapping, path planning and controls have been detailed.

IIT Bombay Racing Driverless: Autonomous Driving Stack for Formula Student AI

TL;DR

The paper presents IIT Bombay Racing's Driverless stack for Formula Student AI, addressing autonomous racing with a sensor-rich architecture that fuses LiDAR and stereo/monocular vision for cone perception, EKF-SLAM for localization, and Delaunay-based path planning with Stanley and Pure Pursuit control. A three-tier perception pipeline integrates LiDAR depth, monocular cues, and stereo depth to handle diverse scenarios, while EKF-SLAM (with MRPT) provides robust localization and mapping, and the Delaunay-based planner yields a racing line with reduced curvature. The system is implemented in ROS2/CAN, validated on a scaled robot and FSDS/EUFS simulators, and demonstrates reliable performance across perception, localization, and control tasks, including edge-case handling like occlusions and fallen cones. The work offers a practical, reliable autonomous racing stack for Formula Student AI events with clear pathways for further robustness in real-world edge cases.

Abstract

This work presents the design and development of IIT Bombay Racing's Formula Student style autonomous racecar algorithm capable of running at the racing events of Formula Student-AI, held in the UK. The car employs a cutting-edge sensor suite of the compute unit NVIDIA Jetson Orin AGX, 2 ZED2i stereo cameras, 1 Velodyne Puck VLP16 LiDAR and SBG Systems Ellipse N GNSS/INS IMU. It features deep learning algorithms and control systems to navigate complex tracks and execute maneuvers without any human intervention. The design process involved extensive simulations and testing to optimize the vehicle's performance and ensure its safety. The algorithms have been tested on a small scale, in-house manufactured 4-wheeled robot and on simulation software. The results obtained for testing various algorithms in perception, simultaneous localization and mapping, path planning and controls have been detailed.
Paper Structure (24 sections, 19 figures, 1 table)

This paper contains 24 sections, 19 figures, 1 table.

Table of Contents

  1. Introduction
  2. Perception
  3. Bounding Box Estimation
  4. LiDAR - Camera Fusion Pipeline
  5. Monocular Vision Pipeline
  6. Using Bounding Box Dimensions
  7. Using Perspective Transformations
  8. Stereo Vision Pipeline
  9. Keypoint Detection
  10. PnP Transform & Feature Matching
  11. Perception Results & Performance
  12. Simultaneous Localisation and Mapping
  13. EKF SLAM
  14. Motion Update
  15. Measurement Update
  16. ...and 9 more sections

Figures (19)

  • Figure 1: High level autonomous system architecture
  • Figure 2: The perception subsystem and parallel use of 3 different pipelines
  • Figure 3: Inference by YOLOv5s. Cones are robustly detected even in presence of noise and blockages
  • Figure 4: Comparision of LiDAR Point Cloud before and after clustering and ground removal. It can be seen that only points on the cones remain
  • Figure 5: PnP formula with extrinsic and camera intrinsic transformation matrices to project LiDAR points onto the image frame b9.
  • ...and 14 more figures