Motion Planning Networks

Adversarial Imitation Via Variational Inverse Reinforcement Learning

Deep reinforcement learning-basedskill transfer and composition method

Multimodal Deep Attention Recurrent Q-Network for perceivable social human-robot interaction.

Artificial potential fields based directionalized sampling for RRT*

A bidirectional version of rrtstar.

Multimodal Deep Q-Network (MDQN) for modelling human-like social intelligence.

This project uses Pepper robot for human tracking, RGBD data acquisition and physical interaction with people.

Human gesture recognition from videos sequences.