A comparison of controller architectures and learning mechanisms for arbitrary robot morphologies

TL;DR

This paper compares different controller and learning method combinations for modular robots with unknown morphologies, finding that a hybrid approach outperforms traditional methods in robustness and efficiency.

Contribution

It introduces a comparative experimental analysis of three controller-learner combinations, highlighting the effectiveness of a hybrid neural network and evolutionary algorithm approach.

Findings

Hybrid controller-learner outperforms traditional methods

The in-between approach is more robust and efficient

Surprising results challenge conventional choices in robot control

Abstract

The main question this paper addresses is: What combination of a robot controller and a learning method should be used, if the morphology of the learning robot is not known in advance? Our interest is rooted in the context of morphologically evolving modular robots, but the question is also relevant in general, for system designers interested in widely applicable solutions. We perform an experimental comparison of three controller-and-learner combinations: one approach where controllers are based on modelling animal locomotion (Central Pattern Generators, CPG) and the learner is an evolutionary algorithm, a completely different method using Reinforcement Learning (RL) with a neural network controller architecture, and a combination `in-between' where controllers are neural networks and the learner is an evolutionary algorithm. We apply these three combinations to a test suite of modular robots and compare their efficacy, efficiency, and robustness. Surprisingly, the usual CPG-based and RL-based options are outperformed by the in-between combination that is more robust and efficient than the other two setups.