Guided Policy Search as Approximate Mirror Descent
This work addresses stability and convergence issues in reinforcement learning for robotics, offering a simpler and more theoretically grounded approach, though it is incremental as it builds on existing guided policy search frameworks.
The paper tackled the problem of improving convergence guarantees and stability in guided policy search algorithms for optimizing complex policies like deep neural networks, showing that these algorithms can be interpreted as approximate mirror descent and deriving a new method with bounded error and empirical performance comparable to or better than prior methods on robotic tasks.
Guided policy search algorithms can be used to optimize complex nonlinear policies, such as deep neural networks, without directly computing policy gradients in the high-dimensional parameter space. Instead, these methods use supervised learning to train the policy to mimic a "teacher" algorithm, such as a trajectory optimizer or a trajectory-centric reinforcement learning method. Guided policy search methods provide asymptotic local convergence guarantees by construction, but it is not clear how much the policy improves within a small, finite number of iterations. We show that guided policy search algorithms can be interpreted as an approximate variant of mirror descent, where the projection onto the constraint manifold is not exact. We derive a new guided policy search algorithm that is simpler and provides appealing improvement and convergence guarantees in simplified convex and linear settings, and show that in the more general nonlinear setting, the error in the projection step can be bounded. We provide empirical results on several simulated robotic navigation and manipulation tasks that show that our method is stable and achieves similar or better performance when compared to prior guided policy search methods, with a simpler formulation and fewer hyperparameters.