Research Projects

This project aims to develop manipulation techniques for bilateral teleoperation of dual arm systems for bimanual tasks, which are essential in various applications such as industrial maintenance, assistive technologies, and disaster response. The project focuses on designing manipulation algorithms using optimal control and reinforcement learning techniques. The lower level controllers for robotic arms are designed using analytical control techniques, while the teleoperation algorithms are designed using reinforcement learning techniques. The segregation of design techniques is based on the availability of simple mathematical models and the presence or absence of possible uncertainties or perturbations. Optimal control guarantees minimum performance levels irrespective of the environment or other exogenous factors, while reinforcement learning-based techniques improve performance based on prior experience. Safety features like collision detection/avoidance and impact minimization will be implemented to ensure safe co-working space for humans and robots. Control algorithms will be validated and reinforcement learning models will be trained on a human-in-the-loop simulation set-up using ROS and haptic devices. An experimental test-bed will be developed with robotic arms and haptic devices for final evaluation and demonstrations. The manipulation techniques developed in this project are expected to be useful for bimanual teleoperation applications like assistive robots, industrial maintenance, rehabilitation, and possibly in high precision applications like surgical robots.