Continual Reinforcement Learning deployed in Real-life using Policy Distillation and Sim2Real Transfer
This work addresses continual learning for robots, but it is incremental as it builds on existing methods like policy distillation and sim2real transfer.
The paper tackled the problem of enabling a robot to learn tasks sequentially without forgetting past ones, using reinforcement learning with state representation learning and policy distillation on 2D navigation tasks, resulting in improved sample efficiency and a single policy that solves all tasks.
We focus on the problem of teaching a robot to solve tasks presented sequentially, i.e., in a continual learning scenario. The robot should be able to solve all tasks it has encountered, without forgetting past tasks. We provide preliminary work on applying Reinforcement Learning to such setting, on 2D navigation tasks for a 3 wheel omni-directional robot. Our approach takes advantage of state representation learning and policy distillation. Policies are trained using learned features as input, rather than raw observations, allowing better sample efficiency. Policy distillation is used to combine multiple policies into a single one that solves all encountered tasks.