IPAB Seminar - 11/05/2023

Title:  Back to the Future: Cognitive Architecture for Robot Manipulation and Control

Abstract:  In this talk, I will begin by providing an overview of an integrated cognitive architecture for knowledge representation, reasoning, control, and learning in robotics. The architecture computationally encodes the principles of step-wise refinement, cognitive systems, and ecological rationality, and leverages the complementary strengths of knowledge-based and data-driven reasoning and learning methods. I will then illustrate the capabilities of this architecture in the context of two robot manipulation and control problems. I will first focus on changing-contact robot manipulation tasks that require the robot to make and break contacts with objects and surfaces. I will describe how our architecture enables a robot to incrementally update its prediction of contact changes, efficiently learn models of the piecewise-continuous interaction dynamics of such tasks, and provide smooth and accurate trajectory tracking based on a variable impedance controller. Next, I will focus on adaptive control of an upper-limb prosthesis using a combination of muscle-tendon units and a variable impedance controller. I will illustrate how our architecture enables a human user to simultaneously control the kinematics, stiffness, and damping of a given degree of freedom, thus supporting adaptation to unexpected external perturbations. If time permits, I will discuss how insights obtained by exploring these robot control problems can help adapt the architecture to address other challenges in robotics.

Speaker bio:  Dr. Mohan Sridharan is a Reader in Cognitive Robot Systems in the School of Computer Science at the University of Birmingham (UK). Prior to his current appointment, he held academic positions at The University of Auckland (NZ) and at Texas Tech University (USA). Dr. Sridharan received his Ph.D. from The University of Texas at Austin (USA). His research interests include knowledge representation, cognitive systems, control systems, and machine learning, as applied to robots and agents collaborating with humans. He is also interested in developing algorithms for automation in non-robotics domains such as climate informatics, agriculture, and transportation.

Web page: http://www.cs.bham.ac.uk/~sridharm