Planning with Goal-Conditioned Policies

Abstract

Planning methods can solve temporally extended sequential decision making problems by composing simple behaviors. However, planning requires suitable abstractions for the states and transitions, which typically need to be designed by hand. In contrast, model-free reinforcement learning (RL) can acquire behaviors from low-level inputs directly, but often struggles with temporally extended tasks. Can we utilize reinforcement learning to automatically form the abstractions needed for planning, thus obtaining the best of both approaches? We show that goal-conditioned policies learned with RL can be incorporated into planning, so that a planner can focus on which states to reach, rather than how those states are reached. However, with complex state observations such as images, not all inputs represent valid states. We therefore also propose using a latent variable model to compactly represent the set of valid states for the planner, so that the policies provide an abstraction of actions, and the latent variable model provides an abstraction of states. We compare our method with planning-based and model-free methods and find that our method significantly outperforms prior work when evaluated on image-based robot navigation and manipulation tasks that require non-greedy, multi-staged behavior.

In Advances in Neural Information Processing Systems, 2019

Cited by in corpus (17)

• RODE: Learning Roles to Decompose Multi-Agent Tasks
• PlanGAN: Model-based Planning With Sparse Rewards and Multiple Goals
• Model-Based Visual Planning with Self-Supervised Functional Distances
• Hierarchical Planning for Long-Horizon Manipulation with Geometric and Symbolic Scene Graphs
• C-Learning: Learning to Achieve Goals via Recursive Classification
• Divide-and-Conquer Monte Carlo Tree Search For Goal-Directed Planning
• Solving Compositional Reinforcement Learning Problems via Task Reduction
• TempoRL: Learning When to Act
• Deep Reactive Planning in Dynamic Environments
• C-Planning: An Automatic Curriculum for Learning Goal-Reaching Tasks
• Adaptable Agent Populations via a Generative Model of Policies
• Hierarchical Robot Navigation in Novel Environments using Rough 2-D Maps
• Self-supervised Visual Reinforcement Learning with Object-centric Representations
• Goal-conditioned Batch Reinforcement Learning for Rotation Invariant Locomotion
• Manipulator-Independent Representations for Visual Imitation
• Learning Synthetic to Real Transfer for Localization and Navigational Tasks
• Hierarchical and Partially Observable Goal-driven Policy Learning with Goals Relational Graph