I have a number of robotic based CS39440/CHM9360/CHM6960 project ideas. Whilst I cannot supervise directly, I am happy to discuss my ideas and act in the role of “on-site customer” with another member of staff (Mark Neal would be most appropriate) acting as the primary supervisor.
An outline of some of my project ideas are listed here. If anyone is interested, please get in touch with me to discuss further.
Learning to Drive
This project will be suitable for a strong candidate on the AI and Robotics degree scheme. The proposed project involves the development of several learning algorithms for the Argo robotic system, the aim of which is to learn to perform turns to specific headings through the adjustment of throttle and steering actuators. Many parameters affect the control of the Argo given its skid-steer nature, and the relatively ‘simple’ task of steering requires complex control strategies that can overcome vibrations and noisy sensor data, inaccuracies of skid-steering and response times of petrol engines. Any control system will need to be able to adapt to and compensate for the changing behaviour of the robotic platform as the surfaces differ. Previous projects have included the development of more traditional, engineered techniques such as PI control. A comparison of the performance of the learning algorithms against standard engineered techniques, including the adaptive potential of the algorithms as surfaces and conditions change, would offer the potential for a significant conference or journal publication for those students intending to pursue a PhD after their undergraduate or masters degree.