Open-source software

  • CNav: C++ code for distributed coordination in multi-agent navigation settings
  • SocialVAE: Python library for state-of-the-art agent trajectory prediction that relies on social attention and a timewise variational autoencoder
  • Implicit Crowds: C++ library for robust and guaranteed collision-free crowd simulation that relies on optimization-based implicit integration
  • ICCGAN: Python code for interactive control of physically simulated characters using reinforcement learning and Generative Adversarial Networks
  • KDMA: Python code for human-inspired multi-agent navigation using reinforcement learning and knowledge distillation
  • NH-TTC: C++library for fast, anticipatory steering of mobile robots having arbitrary equations of motions
  • PFPN: General deep RL framework for particle-based exploration of high-dimensional action spaces during training of physics-based character controllers
  • PowerLaw: C++and Python code for human-like collision avoidance between agents elaborated from analyzing a large corpus of pedestrian trajectory data