Learning Skills to Navigate without a Master: A Sequential Multi-Policy Reinforcement Learning Algorithm

Solving complex problems using reinforcement learning necessitates breaking down the problem into manageable tasks, either explicitly or implicitly, and learning policies to solve these tasks. These policies, in turn, have to be controlled by a master policy that takes high-level decisions. This requires a training algorithm to take this hierarchical decision structure into account when learning these policies. However, training such methods in practice may lead to poor generalization, with either sub-policies executing actions for too few time steps or devolving into a single policy altogether. In our work, we introduce an alternative approach to learn such skills sequentially without using an overarching hierarchical policy. We propose this method in the context of environments where a major component of the objective of a learning agent is to prolong the episode for as long as possible. We refer to our proposed method as Sequential Soft Option Critic. We demonstrate the utility of our approach on navigation and goal-based tasks in a flexible simulated 3D navigation environment that we have developed. We also show that our method outperforms prior methods such as Soft Actor-Critic and Soft Option Critic on our environment, as well as the Gym-Duckietown self-driving car simulator and the Atari River Raid environment.