Level: MSc

Description:

This research investigates evolved controllers for damage recovery on the MechatronicSystems.Group (MS.G) hexapod robot, a spider-like robotic platform that has undergone several development iterations and is currently used for gait optimization studies. Autonomous robots deployed in unstructured or hazardous environments must remain functional under unexpected hardware failures. This project focuses on evaluating already existing different evolved controller encodings, originally developed and tested in 3D physics-based simulations, to assess their effectiveness in enabling adaptive locomotion after leg damage. By transferring simulated controllers to the physical MS.G hexapod, the study seeks to validate simulation-based findings and advance methods for resilient robotic locomotion.

Objectives

• Controller Evaluation – Implement and test a range of evolved controller encodings on the physical MS.G hexapod, including:
  • Compositional Pattern Producing Network (CPPN) parameter encodings of an open-loop controller comprising non-linear oscillators
  • CPPN connection weight encodings of ANN controllers
  • ANN controllers with evolved hidden-layer topology and weights
  • CPPN encodings of Single Unit Pattern Generators (SUPGs)
• Damage Recovery Testing – Examine the performance of each controller under varying levels of leg damage to assess adaptability and robustness.
• Simulation-to-Reality Validation – Demonstrate the transferability of adaptive gaits observed in simulation to real-world experiments on the physical hexapod.
• Comparative Analysis – Compare controller encodings in terms of adaptability, resilience, efficiency, and gait stability in damaged and undamaged conditions.
• Contribution to Resilient Robotics – Provide insights into how evolutionary approaches to controller design can enhance autonomy and robustness in legged robots operating in uncertain environments.

Key skills/interests:

Evolutionary Algorithms & AI, Control systems, Simulation, Programming, Embedded Systems & Hardware Integration, Data Analysis & Experimentation, Robotics & Mechatronics, Bio-inspired Locomotion, Autonomous and Resilient Robotics, Machine Learning for Robotics.

Expected outputs:

1x MSc (Eng), 1x Journal/Conference paper in a leading publication.

Supervisors:

Leanne Raw Associate Professor Geoff Nitschke

Eligibility:

• Full-time student working on campus
• Undergraduate GPA of at least 70%
• Able to begin studies in February 2026