LIDAR Detective

Autonomous vehicle behaviors using object detection with LIDAR

Take a look at a more detailed article here about this project: https://guitar.ucsd.edu/maeece148/index.php/2021FallTeam2

LIDAR Detective is a reinforcement learning system for LiDAR-based autonomous navigation built from scratch for the CUDA-enabled Jetson Nano SBC. By utilizing the CUDA cores found on the Jetson Nano, adaptive autonomous behavior can be enabled with a real-time 100Hz steering control loop. The result is a blazingly fast intelligent control system enabling obstacle avoidance, target following, and other behaviors using a single low-cost LiDAR sensor.

File Structure

• hdr - Header files
• src - Source files

Prerequisites

1. pthread
2. udev
3. CUDA
4. i2c

Prerequisites can be installed on Ubuntu using the following command: apt -y install libudev-dev libpthread-stubs0-dev nvidia-cuda-toolkit libi2c-dev i2c-tools

Build Instructions

1. Create new build directory in project root
2. Navigate to newly created directory
3. Run cmake ..
4. Run make
5. Built files will be found in build directory