Zoom Meeting: https://umass-amherst.zoom.us/j/91028501426?pwd=S3F0cHhMNm1DbkVPMzJQRGoxbHpTQT09
Meeting ID: 910 2850 1426 Passcode: 314159
Metareasoning is the process by which an autonomous system optimizes its own planning and execution processes in order to operate more effectively in its environment. As autonomous systems grow in sophistication and autonomy, the need for metareasoning has become critical for efficient and reliable operation in noisy, stochastic, unstructured domains for long durations. However, despite considerable progress in metareasoning as a whole over the last thirty years, techniques that improve planning rely on many assumptions that diminish accuracy and practical utility while those that improve execution have not seen much attention yet. The main goal of this thesis is to propose metareasoning approaches that improve both the efficiency of planning and the reliability of execution in autonomous systems.
This thesis therefore proposes a two-pronged framework that offers more effective metareasoning for planning and expands the scope of metareasoning to execution. We begin by offering two metareasoning approaches for efficient planning: (1) a method that determines the optimal stopping point of an anytime planner by predicting performance at runtime and (2) a method that estimates the optimal hyperparameters of an anytime planner by learning from experience with deep reinforcement learning. We then offer two metareasoning approaches for reliable execution: (3) a method that recovers from exceptional circumstances by interleaving a main decision process with exception handlers and (4) a method that maintains and restores safe operation by running a main decision process and safety processes in parallel using a decision-theoretic algorithm for conflict resolution.
Advisor: Shlomo Zilberstein