Research

  • Autonomous robotic manipulation by data-efficient learning.
  • Deep reinforcement learning of dynamic locomotion of legged robots.
  • Meta-learning of sensorimotor skills of robots.
  • Multi-contact motion planning of multi-limbed robots.
  • Immersive tele-operation with multi-sensory interface.
  • Autonomous stair-climbing of delivery robot.
  • Transfer of human-inspired motor control.
  • Bio-inspired robotic gripper for handling soft objects.
  • Research Outcomes

    Autonomous fall recovery

    Deep reinforcement learning to acquire whole-body motor skills: standing-up policies for legged robots. The learning algorithm can successfully discover various recovery behaviors by exploiting whole-body contacts with the environment to achieve this dynamic and contact-rich task.

    Learning obstacle negotiation and fall recovery for mobile robot

    Anthropomorphic robotic hand manipulation and grasping

    Haptic tele-manipulation

    Haptic-guided tele-operation using Sigma.7 and Franka arm: sense interaction forces and exploit affordance.

    Autonomous robotic manipulation: reactive grasping

    Deep reinforcement learning for autonomous grasping: learned re-grasping behavior to handle slippage and disturbances.

    Fine control of grasping force

    Compliant force control for strawberry-picking.

    Immersive virtual reality for manipulation

    Immersive multimodal interface for controlling Shadow hands to handle and manipulate various objects: immersive VR headset, optical markerless hand-tracking and haptic gloves.

    Object detection and semantic segmentation

    Robot perception for robot manipulation tasks such as grasping.

    Robust walking and push recovery

    Optimisation and predictive control to counterbalance external pushes.

    Stabilization and dynamic walking of humanoids

    Unified formulation of Model Predictive Control of legged robots: for both offline gait planning and online feedback control. The method is able to overcome unmodeled pushes and terrain irregularities.

    Contribution in past EU projects

  • WALK-MAN (FP7-ICT 611832) Research Role: robust balancing capability and locomotion control that allow humanoid robots to walk in cluttered and unstructured environment.
  • AMARSi (FP7-ICT 248311) Research Role: compliance control of novel mechanical systems with passive compliance for safer human robot interaction, and development of control solutions for better gait stabilization of COmpliant HuMANoid robot (COMAN).