Thresholded Lexicographic Ordered Multiobjective Reinforcement Learning
This addresses lexicographic tasks in reinforcement learning, which are common in real-life scenarios but have limited existing work, though the approach appears incremental as it builds on prior heuristics.
The paper tackles the problem of lexicographic multi-objective reinforcement learning, where objectives have a strict importance order, by proposing fixes for practical performance issues and introducing a policy optimization approach called Lexicographic Projection Optimization (LPO). The result is demonstrated on benchmark problems, though no concrete performance numbers are provided in the abstract.
Lexicographic multi-objective problems, which impose a lexicographic importance order over the objectives, arise in many real-life scenarios. Existing Reinforcement Learning work directly addressing lexicographic tasks has been scarce. The few proposed approaches were all noted to be heuristics without theoretical guarantees as the Bellman equation is not applicable to them. Additionally, the practical applicability of these prior approaches also suffers from various issues such as not being able to reach the goal state. While some of these issues have been known before, in this work we investigate further shortcomings, and propose fixes for improving practical performance in many cases. We also present a policy optimization approach using our Lexicographic Projection Optimization (LPO) algorithm that has the potential to address these theoretical and practical concerns. Finally, we demonstrate our proposed algorithms on benchmark problems.