Robotics and AI are identified as key growth areas across the world. The robots of today are no longer confined to structured environments, nor are they completely isolated from humans. Biological systems (humans/animals) naturally exhibit energy-eficient, robust, and adaptive behaviours in complex and contact-rich environments, whilst the existing robotic systems are still sufering from insuficient capabilities of sensory-motor and learning. Humans can perform a range of tasks with planning and excellence, but these are very dificult for robots. Motion learning and learning are the analysis of and planning for objects moving through space. It is a part of research problems across disciplines and requires specialised treatments. Motion planning algorithms move robots safely through complicated environments, validate both the assembly and operation of multipart systems, and solve a variety of other tasks. Real-time robot motion planning has become an active yet challenging research area recently, particularly with the issues of modelling of environmental interactions, recognition and grasping of deformable objects and optimization in contact-rich scenarios. In this presentation, I will start with the needs and challenges of autonomous motion learning in contact-rich and uncertain scenarios, and then explore plausible solutions to approach these problems with some interesting applications in agriculture and manufacturing.
Biography