Multi-robot Social Navigation

We are working on developing a deep reinforcement learning (DRL) based social navigation approach for multiple intelligent robots to safely move in pedestrian-rich environments via cooperative perception. For simulation experiments, we developed multiple ROS Gazebo simulation scenarios rich in social behaviors of pedestrians using social force model. In order to enable robot perception for pedestrians and trajectories, we are planning to apply 3D multi-object tracking algorithms for Cooperative Perception using 3D Lidar raw sensor data.

The project is on work, please stay tuned and keep updated.

The simulation of social crowds using social force model (Pedsim_ROS)