Harish K. Venkataraman

Harish Venkataraman, Derya Aksaray, Peter Seiler

Signal temporal logic (STL) is an expressive language to specify time-bound real-world robotic tasks and safety specifications. Recently, there has been an interest in learning optimal policies to satisfy STL specifications via reinforcement learning (RL). Learning to satisfy STL specifications often needs a sufficient length of state history to compute reward and the next action. The need for history results in exponential state-space growth for the learning problem. Thus the learning problem becomes computationally intractable for most real-world applications. In this paper, we propose a compact means to capture state history in a new augmented state-space representation. An approximation to the objective (maximizing probability of satisfaction) is proposed and solved for in the new augmented state-space. We show the performance bound of the approximate solution and compare it with the solution of an existing technique via simulations.

Harish K. Venkataraman, Peter J. Seiler

Reinforcement learning (RL) is used to directly design a control policy using data collected from the system. This paper considers the robustness of controllers trained via model-free RL. The discussion focuses on the standard model-based linear quadratic Gaussian (LQG) problem as a special instance of RL. A simple example, originally formulated for LQG problems, is used to demonstrate that RL with partial observations can lead to poor robustness margins. It is proposed to recover robustness by introducing random perturbations at the system input during the RL training. The perturbation magnitude can be used to trade off performance for robustness. Two simple examples are presented to demonstrate the proposed method for enhancing robustness during RL training.