Composable Learning with Sparse Kernel Representations
This addresses the challenge of sample efficiency and policy composition in robotics, though it appears incremental as it builds on existing kernel and advantage function methods.
The paper tackles the problem of learning sparse non-parametric controllers in reinforcement learning by using a normalized advantage function to improve sample complexity and enable efficient composition of policies without extra training. The result demonstrates successful obstacle-avoidance policies in simulations and physical transfer, with composed policies retaining component performance.
We present a reinforcement learning algorithm for learning sparse non-parametric controllers in a Reproducing Kernel Hilbert Space. We improve the sample complexity of this approach by imposing a structure of the state-action function through a normalized advantage function (NAF). This representation of the policy enables efficiently composing multiple learned models without additional training samples or interaction with the environment. We demonstrate the performance of this algorithm on learning obstacle-avoidance policies in multiple simulations of a robot equipped with a laser scanner while navigating in a 2D environment. We apply the composition operation to various policy combinations and test them to show that the composed policies retain the performance of their components. We also transfer the composed policy directly to a physical platform operating in an arena with obstacles in order to demonstrate a degree of generalization.