We conduct research and development revolving around the area of robot dexterity in blurring the boundary between humans and service robots.
Our Dexterous Robotics(Dexbot) Research is currently organised into four research themes:
Crossmodal Tactile Perception
The sense of touch is arguably the first developed human sense in its infancy. It grounds other sensory systems and acts as a last line of defence during a physical interaction with the environment. However, unlike other electronic sensors such as camera and microphone, representation of the spatio-temporal tactile sensory data varies across different manufacturers making it a challenging task for large-scale perceptual learning.
We seek to design and implement effective platform-agnostic learning paradigms that enrich the understanding of the environment during interactions with the sense of touch. Such crossmodal understanding provides timely and accurate perceptual information needed for safe and fine dexterous manipulation of the environment.
Click here for sample publications.
Click here for sample publications.
Adaptive & Dexterous Control
Apart from mobility, our daily activities are largely carried out by our hands(or end-effectors in the robotics terms). In the presence of constant environmental disturbances, we can perform grasping, in-hand manipulations, bimanual operations or even social touches almost effortlessly by synergetically controlling more than 30 muscles (or actuators) on each hand. This is made possible by the adaptive and dexterous motor control loops of the motor cortex and the spinal cord. These control loops largely depend on the external stimuli presented to our sensory systems.
This research theme seeks to construct and implement task-specific adaptive control systems that are capable of performing real-world dexterous operations incorporating multimodal sensory information into the hierarchy of the control loop.
Click here for sample publications.
Click here for sample publications.
Robot Skills Learning
Robots to provide service in the human space are expected to possess an inexhaustible skills to suit the different application scenario. However, programming these skills by hand is a labourious effort. The aim of this research theme is to grow the repertoire of embodied dexterous skills for complex robotic systems with high number of degrees of freedom (DoFs) with the use of machine learning techniques. The robots are to pick up these skills through self-exploration, expert demonstration or a combination of both.
Click here for sample publications.
Click here for sample publications.
Human-robot Interaction
As robotics technologies mature, robots are expected to work in a team and collaborating with other people, not only in a factory setup as a cobot but also increasing in our living spaces.
Focusing on the use of learning techniques, multiple sensory and interaction modalities, we seek to investigate on:
Click here for sample publications.
- Methods for safe interactions
- Methods for (socially) compliant and collaborative interactions
- Methods for coordinated actions to improve the quality of task completion
- Co-working and buddy robots for the professional industry
- Assistive robots to augment human potentials
- Care robots to improve quality of life
Click here for sample publications.