IPAB Workshop-29/07/2021

Joao Pousa De Moura

Title: Non-prehensile Manipulation with Non-Linear MPC

Abstract: Manipulating objects without fully grasping them, known as non-prehensile manipulation, can be quite challenging due to the complex nature of the contact interaction between the robot and the manipulated object.

Recently, I have started studying this type of contact interactions, with special emphasis on sliding contacts, and in this talk I will go through a case study of how to plan and control the motion of a planar object using a non-linear Model Predictive Controller (MPC).

Chenyang Zhao

Title: Robust Domain Randomised Reinforcement Learning through Peer-to-Peer Distillation

Abstract: In reinforcement learning, domain randomisation is a popular technique for learning general policies that are robust to new environments and domain-shifts at deployment. However, naively aggregating information from randomised domains may lead to high variance in gradient estimation and an unstable learning process. To address this issue, we present a peer-to-peer online distillation strategy for reinforcement learning termed P2PDRL, where multiple learning agents are each assigned to a different environment, and then exchange knowledge through mutual regularisation based on Kullback–Leibler divergence. Our experiments on continuous control tasks show that P2PDRL enables robust learning across a wider randomisation distribution than baselines, and more robust generalisation to new environments at testing.