Reinforcement Learning with Symbolic Reward Machines

Researchers propose Symbolic Reward Machines (SRMs) as an improvement over traditional Reward Machines in reinforcement learning, eliminating the need for manual user input while maintaining performance. SRMs process observations directly through symbolic formulas, making them more applicable to widely adopted RL frameworks.

• →Symbolic Reward Machines overcome limitations of traditional Reward Machines by removing the need for manual labeling functions.
• →SRMs process standard environment output directly through guards represented by symbolic formulas.
• →The proposed QSRM and LSRM algorithms outperform baseline RL approaches while matching existing RM method results.
• →SRMs provide interpretable task representations while adhering to widely used environment definitions.
• →This advancement improves applicability of reward machines in mainstream reinforcement learning frameworks.