CrowdSusrfer on Husky(without Global Plan)

Published:

CrowdSurfer is a Path Planning Algorithm that is built on top of the Sampling based optimizer called PRIEST(our previous work). CrowdSurfer makes use of a Vector Quantized VAE to generate a distribution diverse motion primitives for PRIEST to sample from. The VQVAE model is trained by Behavior Cloning using Expert Demonstrations of a wheelchair navigating in extremely crowded environments.

This work has been submitted to IEEE ICRA 2025 and is currently under review.

Features

  • Multi-modal in nature.
  • Can navigate in extremely crowded environments with more than 50 humans.
  • Excellent run-time speeds of 40ms per trajectory rollout.

Technologies Used

  • Generative Modeling (VQ-VAE + PixelCNN)
  • Batch Optimization
  • ROS Nav Stack

Link to Project Repository
View the project on YouTube