Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper compares Central Pattern Generators (CPGs) and Multi-Layer Perceptrons (MLPs) for robot control, focusing on the impact of parameter space size. Through evolutionary and reinforcement learning on a robot morphology with limited proprioception, they found that shallow MLPs and densely connected CPGs outperform deeper MLPs and Actor-Critic architectures. They introduce a Parameter Impact metric, revealing that the increased parameters of reinforcement learning do not yield superior performance, thus favoring evolutionary strategies in this context.
More parameters don't always mean better robot control: shallow networks can outperform deep RL in tasks with limited input/output spaces.
While Central Pattern Generators (CPGs) and Multi-Layer Perceptrons (MLP) are widely used paradigms in robot control, few systematic studies have been performed on the relative merits of large parameter spaces. In contexts where input and output spaces are small and performance is bounded, having more parameters to optimize may actively hinder the learning process instead of empowering it. To empirically measure this, we submit a given robot morphology, with limited proprioceptive capabilities, to controller optimization under two bio-inspired paradigms (CPGs and MLPs) with evolutionary- and reinforcement- trainer protocols. By varying parameter spaces across multiple reward functions, we observe that shallow MLPs and densely connected CPGs result in better performance when compared to deeper MLPs or Actor-Critic architectures. To account for the relationship between said performance and the number of parameters, we introduce a Parameter Impact metric which demonstrates that the additional parameters required by the reinforcement technique do not translate into better performance, thus favouring evolutionary strategies.