Program Machine Policy: Addressing Long-Horizon Tasks by Integrating Program Synthesis and State Machines
This work addresses the problem of scalability and generalization in reinforcement learning for long-term tasks, representing an incremental improvement over existing programmatic and state machine methods.
The paper tackles the challenge of addressing long-horizon tasks in reinforcement learning by integrating program synthesis and state machines, resulting in a framework that outperforms baselines and generalizes to longer horizons without fine-tuning.
Deep reinforcement learning (deep RL) excels in various domains but lacks generalizability and interpretability. On the other hand, programmatic RL methods (Trivedi et al., 2021; Liu et al., 2023) reformulate RL tasks as synthesizing interpretable programs that can be executed in the environments. Despite encouraging results, these methods are limited to short-horizon tasks. On the other hand, representing RL policies using state machines (Inala et al., 2020) can inductively generalize to long-horizon tasks; however, it struggles to scale up to acquire diverse and complex behaviors. This work proposes the Program Machine Policy (POMP), which bridges the advantages of programmatic RL and state machine policies, allowing for the representation of complex behaviors and the address of long-term tasks. Specifically, we introduce a method that can retrieve a set of effective, diverse, and compatible programs. Then, we use these programs as modes of a state machine and learn a transition function to transition among mode programs, allowing for capturing repetitive behaviors. Our proposed framework outperforms programmatic RL and deep RL baselines on various tasks and demonstrates the ability to inductively generalize to even longer horizons without any fine-tuning. Ablation studies justify the effectiveness of our proposed search algorithm for retrieving a set of programs as modes.