Dom is a final year PhD Student in Intelligent Automation, researching how to teach robots to carry out complex manipulation tasks. Previous to this post, Dom worked within the university's Engines and Powertrains Research Lab, where he was responsible for instrumenting research engines and developing tools to collect data and control engine parameters. This work often involved the design and manufacture of bespoke electrical and mechanical parts and devices made to a very high standard.
Dom completed his 2:1 BSc (Hons) degree in Electrical and Electronic Engineering in 2017. Dom's final year project was working with the Control Systems Group at Loughborough University where he developed a novel safe robotic actuator using active compliant control methods. The actuator was capable of sensing external interaction forces less than 0.2 newtons without the use of a force/torque sensor.
PhD Thesis Title: Learning Complex Manipulation Dynamics
Robots can perform a large variety of task in industry, but one task has eluded automation due to its complex nature. Fitting long flexible objects, such as wiring looms, seals and hoses requires great dexterity, but also, an appreciation for the interaction dynamics between the object and its environment.
Threading flexible objects through holes in a chassis is a relatively simple task for a human, but a robot does not know how manipulating one end of an object in free space may change those contact dynamics at the other end. Reinforcement learning, a branch of machine learning, may provide a learning solution; however, it's not clear if reinforcement learning will be able to cope with changes to the system dynamics when changes are made to previously visited states by the flexible object which is still in contact with those past states.