Humanoid robotics research has come a long way in recent years with impressive state estimation, planning, and control algorithms able to synthesise and execute motion online for static cluttered environments. They do face challenges, however, as discrepancies between planned behaviour and control execution occur. Our work aims at increasing the robustness and autonomy of humanoid robots by allowing them to replan their motion online in response to disturbance, environment and contact changes, where behaviours are specified through high-level objective functions. Hereto, the research focuses on optimal and model-predictive control methods targeted at high-dimensional systems in closed loop with perception input. Key challenges in this ﬁeld include multi-contact, stochasticity, and dimensionality which make replanning online in response to sudden changes a challenge for fulfilling real-time requirements. Our approach makes use of offline generated samples to determine a quickly solvable reduced optimal control problem that provides a near-optimal warm-start to the iterative optimisation algorithm.
Publications: Google Scholar Profile
Research Interests: My main research interests are in making robots operate smarter, more autonomously and robust, while exposing less complexity. I have previously worked on
- Transferring human demonstrated motion automatically to robots of different morphologies by leveraging topology and machine learning (Bachelor thesis: Human to Humanoid Robot Motion Transfer using Interaction Meshes and Machine Learning),
- Collaborative robots to assist human workers in shared workcells and environments (internship at the Fraunhofer Institute for Manufacturing Engineering and Automation IPA).
- Shared autonomy powered by dense visual mapping and collision-free motion planning to allow robots to operate in real-world environments for longer and recover from unpredictable changes more easily (Master of Science by Research thesis: Dense Visual Mapping and Planning for Robust Autonomy).
I am currently working on efficient formulations for robust optimal control which allow online replanning in multi-contact scenarios on humanoid robots. My interests hereby are in parallelisation and machine learning to allow algorithms to scale to high-dimensional scenarios and real-world applications. I am a core member of our humanoid robotics team working on the Edinburgh-NASA Valkyrie Project.
Education: I have previously completed a BEng (Hns, 1st Class) in Mechanical Engineering with Management at The University of Edinburgh and The University of Texas at Austin and a MSc by Research in Robotics and Autonomous Systems with Distinction at The University of Edinburgh. I have also completed the Graduate Studies Programme at Singularity University in Silicon Valley.
Notable Experience: I have previously completed worked on industrial robotics applications for small and medium enterprises (co-bots).
At the age of 14, I founded a service robotics research lab in collaboration with the world’s largest manufacturer of industrial robots to develop a kitchen robot, and raised a six digit figure in funding to enable high school students to research and participate in competitions internationally.
As a consultant and former startup founder, I’ve developed and deployed software in government, law enforcement, legal document handling, education, and entertainment with numerous household brands and advise young and quickly growing ventures on consulting and board level.