pylot  by erdos-project

Autonomous vehicle platform for developing and testing self-driving components

created 6 years ago
505 stars

Top 62.5% on sourcepulse

GitHubView on GitHub
Project Summary

Pylot is a modular autonomous driving platform designed for developing and testing vehicle components like perception, prediction, and planning. It targets researchers and engineers working with the CARLA simulator or real-world vehicles, offering a flexible framework to explore latency-accuracy trade-offs.

How It Works

Pylot is built on the ERDOS framework, enabling components to be developed and executed as independent ERDOS operators. This modularity allows for isolated testing of specific functionalities (e.g., obstacle detection, lane detection, planning) or end-to-end execution. The platform supports various algorithms for each component, including different object detection models, planning strategies (waypoint, Frenet, RRT*, Hybrid A*), and control methods (PID, MPC).

Quick Start & Requirements

  • Docker: nvidia-docker run -itd --name pylot -p 20022:22 erdosproject/pylot /bin/bash
  • Prerequisites: NVIDIA Docker, X server forwarding for visualization.
  • Setup: Docker setup is straightforward; manual installation involves running ./install.sh and setting environment variables.
  • Documentation: Setup Instructions, Documentation

Highlighted Details

  • Supports multiple object detection models (frcnn_resnet101, ssd-mobilenet-fpn-640, ssdlite-mobilenet-v2).
  • Offers diverse planning algorithms including Frenet Optimal Trajectory, RRT*, and Hybrid A*.
  • Includes a data collection script for CARLA sensor data.
  • Provides a baseline agent for the CARLA Autonomous Driving Challenge.

Maintenance & Community

Health Check
Last commit

2 years ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
12 stars in the last 90 days

Explore Similar Projects

Feedback? Help us improve.