VACOP - Open-Source, Plug & Play Connected Autonomous Vehicle

The VACOP project (Open-Source, Plug & Play Connected Autonomous Vehicle) aims to develop a connected autonomous navigation platform on the autOCampus campus at IRIT. The goal is to enable the vehicle to navigate safely within a controlled environment, while providing real-time supervision and control through a dedicated web interface and a private 5G infrastructure.

This project is part of IRIT’s research and experimentation activities in autonomous mobility, as well as the SRI 2026 engineering program.

---

Main Features

The VACOP system is designed around four main modules:

• Localization and State Estimation:

  • Centimeter-level geolocation accuracy (2–3 cm) using an RTK GNSS receiver (u-blox ZED-F9P).
  • Robust data fusion using an Extended Kalman Filter, combining GNSS and odometry data from Hall-effect sensors to maintain position estimation even in case of GNSS signal loss.

• Perception and Environment Understanding:

  • SLAM mapping and continuous localization using a 3D LiDAR sensor (Robosense Helios 16P).
  • Detection and classification of static and dynamic obstacles, such as pedestrians and vehicles, through LiDAR and camera data fusion (RGB and RGB-D).
  • Generation of a dynamic costmap for navigation.

• Trajectory Planning:

  • Global planning to compute the optimal route toward a target destination.
  • Local planning for real-time obstacle avoidance and risk-aware adaptation, including collision and pedestrian-related risks.
  • Closed-loop control of actuators, including motors, steering, and braking.

• Communication and Supervision Interface:

  • Web-based supervision interface built with React and Flask for remote control and visualization.
  • Autonomous Mode: Mission planning, either immediate or scheduled, by selecting a destination on the map.
  • Teleoperation Mode: Remote manual control of the vehicle using a gamepad.
  • Real-Time Visualization: Display of video streams, vehicle position, logs, and obstacle maps.

---

System Architecture

The architecture is built on ROS 2 and organized around three main components, connected through a private 5G network.

1. Vehicle — Onboard System:

   • Main computing unit: NVIDIA Jetson Orin NX for sensor fusion, perception, and planning.
   • Sensors: LiDAR, RGB/RGB-D cameras, RTK GNSS, Hall-effect sensors for odometry.
   • Communication: Telit 5G module for telemetry and NTRIP corrections.
   • Low-level control: Raspberry Pi and SOLO MEGA motor controllers.

2. Servers — IRIT Infrastructure:

   • Backend (C2): A Flask application handling business logic, the REST API, WebSockets for real-time data, and authentication.
   • Database (C2): PostgreSQL for storing users, missions, and log history.
   • Deployment: The entire supervision stack (C1 + C2) is containerized with Docker.

3. Client — Operator Interface:

   • Frontend (C1): A secure React web interface using HTTPS, allowing the operator to supervise and control the vehicle.
   • Protocols: HTTPS, WebSocket, and WebRTC for video streaming.