Sim2real gap is non-monotonic with robot complexity for morphology-in-the-loop flapping wing design

TL;DR

This study investigates how the sim2real transfer gap varies with robot complexity in flapping wing design, revealing a non-monotonic relationship that suggests certain morphological details can reduce the reality gap.

Contribution

It introduces a parameterized morphology design space based on biological features and measures the non-monotonic relationship between complexity and sim2real gap in flapping wings.

Findings

The reality gap varies non-monotonically with design complexity.

Certain morphological features can narrow the sim2real gap.

The results suggest potential for automated morphology optimization.

Abstract

Morphology of a robot design is important to its ability to achieve a stated goal and therefore applying machine learning approaches that incorporate morphology in the design space can provide scope for significant advantage. Our study is set in a domain known to be reliant on morphology: flapping wing flight. We developed a parameterised morphology design space that draws features from biological exemplars and apply automated design to produce a set of high performance robot morphologies in simulation. By performing sim2real transfer on a selection, for the first time we measure the shape of the reality gap for variations in design complexity. We found for the flapping wing that the reality gap changes non-monotonically with complexity, suggesting that certain morphology details narrow the gap more than others, and that such details could be identified and further optimised in a future end-to-end automated morphology design process.