Shengyin Wang

Profile

I am a second-year Ph.D. student working at the Robot Manipulation Laboratory with Dr. Mehmet Dogar, in the School of Computing, University of Leeds. Before that, I acquired my bachelor's degree and master’s degree in China. I envision a future of service robots helping in every house and making a comfortable life for every human being especially for those with disabilities or elderly people. In consideration of various tasks and unstructured environments inside a house, we need the robot to be both intelligent and versatile, in order to deal with a wide variety of objects including rigid and deformable ones.

Research interests

Loads work has been done for rigid object manipulation, however, research on deformable object manipulation still lags behind, mainly due to the challenges of infinite-dimensional state space and complex dynamics. To deploy a service robot working within the household environment, I am particularly interested in developing deformable object manipulation methods. The core ideas of my research are model simplification for motion planning and combining deep reinforcement learning with planning to address the two challenges specifically for deformable objects. Although deformable objects have high-dimensional state space, not every particle is relevant to complete the task, thus we are trying to extract the most indispensable element to achieve the goal, which will greatly reduce the computation burden of the motion planning problem. In the future, we will combine deep reinforcement learning with the model simplification-based motion planning scheme we are developing now to exploit the merits of both methods while avoiding their drawbacks.

Qualifications

• MSc (Eng) & BSc