Learning Parameterized Skills
This work addresses the challenge of generalizing skills across varied tasks in robotics, though it appears incremental as it builds on existing manifold learning and regression techniques.
The authors tackled the problem of learning skills that can adapt to a distribution of parameterized reinforcement learning tasks by modeling policy parameters as a lower-dimensional manifold, and demonstrated their method on a simulated robotic arm achieving accurate dart throws to target locations.
We introduce a method for constructing skills capable of solving tasks drawn from a distribution of parameterized reinforcement learning problems. The method draws example tasks from a distribution of interest and uses the corresponding learned policies to estimate the topology of the lower-dimensional piecewise-smooth manifold on which the skill policies lie. This manifold models how policy parameters change as task parameters vary. The method identifies the number of charts that compose the manifold and then applies non-linear regression in each chart to construct a parameterized skill by predicting policy parameters from task parameters. We evaluate our method on an underactuated simulated robotic arm tasked with learning to accurately throw darts at a parameterized target location.