PhD Studentship: Development of a Visuo-Haptic Recognition System to Enhance Robotic Perception in Cluttered Environments

Funding is available to cover full home tuition fees and a stipend at standard UKRI rate (£19,237 per annum for 24/25) for 3.5 years. Start date is September 2024.

The project aligns and contributes to the long-term goal of nuclear decommissioning at RAICo1, Dalton Cumbria and The University of Manchester. During nuclear decommissioning, operators face severe risks from unknown environment to dealing with hazardous and sharp objects such as exposed wires and corroded metal objects. Accurate object recognition even in poorly lit and diverse environments is thus critical to operational safety. By enhancing robots with perception capabilities through visuo-haptic sensing, this project aims to significantly improve safety and efficiency in many decontamination processes.

To enable object recognition in challenging environments, the project will research and develop a visuo-haptic recognition system that combines tactile sensing with a wide range of vision sensors (including visible, infrared, and depth), aiming to enhance and improve the reliability of object recognition and perception. An advanced multi-modal object recognition approach will be developed to integrate and optimally fuse features extracted from both visual and haptic information. To reduce recognition vagueness, active sensing methods may also be utilised, enabling the robot to undertake optimal actions to explore an environment and recognise objects efficiently. Furthermore, we will extract sensory information from operators' manoeuvres, incorporating their insights of environmental assessments in robotic perception. This advanced perception capability is essential for robots to execute critical nuclear decommissioning tasks, such as dismantling, thereby contributing to more efficient, more reliable and safer operational procedures.

Applicants should have at least an upper second-class (2:1) undergraduate degree (or overseas equivalent) in Control Engineering, Electronic Engineering, Mechanical Engineering, Computer Science or related disciplines.

Experience in computer vision, object recognition, machine learning, robotic manipulator control, active sensing, human-robot interaction will be an advantage. Desirable skills include expert programming (C++ and Python), deep learning, experience with ROS and experimental implementation.

Interested applicants are welcome to contact the supervisors before applying: Dr. Xiaoxiao Cheng (xiaoxiao.cheng@manchester.ac.uk) and Prof. Hujun Yin (hujun.yin@manchester.ac.uk)

Advert information