Fine Tuning Swimming Locomotion Learned from Mosquito Larvae

TL;DR

The paper addresses improving a parameterized, bio-inspired swimmer by using RL-guided local search to optimize motion parameters while leveraging a CFD force clone to reduce costly simulations. It demonstrates that Baseline Guided Policy Search can identify beneficial mid-episode perturbations and that a deep neural network can faithfully predict surface forces to enable model-based optimization. Key findings show LSTM-based clones (depth 3) perform best and that the RL-based refinements yield gradual improvements in displacement, albeit modest in scale. These methods offer a path toward efficient, data-driven optimization of locomotion for bio-inspired underwater robots.

Abstract

In prior research, we analyzed the backwards swimming motion of mosquito larvae, parameterized it, and replicated it in a Computational Fluid Dynamics (CFD) model. Since the parameterized swimming motion is copied from observed larvae, it is not necessarily the most efficient locomotion for the model of the swimmer. In this project, we further optimize this copied solution for the swimmer model. We utilize Reinforcement Learning to guide local parameter updates. Since the majority of the computation cost arises from the CFD model, we additionally train a deep learning model to replicate the forces acting on the swimmer model. We find that this method is effective at performing local search to improve the parameterized swimming locomotion.

Figures (5)

• Figure 1: Model of 2D swimmer in CFD
• Figure 2: Local parameter search using BGPS algorithm
• Figure 3: Comparison of log test losses of various types of models
• Figure 4: Comparison of log test losses of various LSTM network depths
• Figure 5: Result of BGPS on swimmer displacement per episode