Hybrid global/local optimisation for the design of diverse structures

The School of EECS is hosting the following HDR Progress Review 1 Confirmation Seminar:

Hybrid global/local optimisation for the design of diverse structures

Speaker: Kurt Enkera

Abstract:

Advanced computational design algorithms have revolutionised industries such as healthcare, aerospace, automotive, and construction by automating the creation of high-performance structures, including aerospace components, medical prosthetics, and vehicle chassis. However, robotics has yet to fully leverage these powerful design tools due to the challenging need to simultaneously optimise both the morphology and control of a robot in dynamic environments. This research introduces a novel hybrid framework that combines two traditionally separate design approaches – evolutionary algorithms (EAs) and topology optimisation (TO) - to create a computational tool for designing mechanically intelligent robotic systems. While EAs are useful for discovering diverse robot morphologies and controllers, the robots they have produced to date are extremely limited. We aim to improve the performance of evolutionary robotics design algorithms in two ways. First, we allow TO to further optimise morphologies discovered by an EA towards global optimality in an automated manner. Secondly, we explore how wavelet-based digital design representations can enable EAs to systematically search previously inaccessible design spaces, offering a path towards more expressive and effective robotic design tools.