Project description:

The MechatronicSystems.Group (MS.G) hexapod robot is a spider-like robot that is the result of multiple generations and versions involving the work of several students over the years. The project is at currently at a mature phase of development and has most recently been used for gait determination work.

The objective of this project is to evaluate various hexapod robot controller encodings on the MS.G hexapod robot given varying degrees of leg damage. The project entails taking controllers evolved in 3D physics-based simulation [1] for various controller encodings, including: a kinematic and trajectory based open-loop controller, a Compositional Pattern Producing Network (CPPN) parameter encoding of an open-loop controller comprising non-linear oscillators, CPPN connection weight encoding of an Artificial Neural Network (ANN), an ANN controller, where hidden-layer topology and connection weights were evolved using an evolutionary algorithm, and a CPPN encoding of a Single Unit Pattern Generator (SUPG). The project goal is to demonstrate the same adaptive gaits (given leg damage) in the physical hexapod, as observed in simulation [2].

[1] Mailer, C., Nitschke, G., and Raw, L. (2021). Evolving Gaits for Damage Control in a Hexapod Robot. In Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2021), pages 146-153, ACM, Lille, France.

[2] Pouroullis, A., Blore, D., Scott, M., Smith, J., Mkhatshwa, S., and Nitschke, G. (2024). Automated Damage Recovery in a Hexapod Robot. ACM Transactions on Evolutionary Learning and Optimization (TELO). Submitted.

Supervisor:

Leanne Raw