Structured Control Nets for Deep Reinforcement Learning

TL;DR

This paper introduces Structured Control Nets, a neural network architecture for deep reinforcement learning that combines linear and nonlinear modules to improve training efficiency, reward, and generalization across various control tasks.

Contribution

The paper proposes a novel neural network architecture, Structured Control Net, that splits the policy network into linear and nonlinear modules to enhance performance and efficiency in reinforcement learning.

Findings

Achieved competitive results on MuJoCo, Roboschool, Atari, and urban driving environments.

Demonstrated improved locomotion performance by emulating biological CPGs.

Validated benefits across multiple training methods and ablation tests.

Abstract

In recent years, Deep Reinforcement Learning has made impressive advances in solving several important benchmark problems for sequential decision making. Many control applications use a generic multilayer perceptron (MLP) for non-vision parts of the policy network. In this work, we propose a new neural network architecture for the policy network representation that is simple yet effective. The proposed Structured Control Net (SCN) splits the generic MLP into two separate sub-modules: a nonlinear control module and a linear control module. Intuitively, the nonlinear control is for forward-looking and global control, while the linear control stabilizes the local dynamics around the residual of global control. We hypothesize that this will bring together the benefits of both linear and nonlinear policies: improve training sample efficiency, final episodic reward, and generalization of learned policy, while requiring a smaller network and being generally applicable to different training methods. We validated our hypothesis with competitive results on simulations from OpenAI MuJoCo, Roboschool, Atari, and a custom 2D urban driving environment, with various ablation and generalization tests, trained with multiple black-box and policy gradient training methods. The proposed architecture has the potential to improve upon broader control tasks by incorporating problem specific priors into the architecture. As a case study, we demonstrate much improved performance for locomotion tasks by emulating the biological central pattern generators (CPGs) as the nonlinear part of the architecture.