Projects

Learning For Control: An Inverse Optimization Approach

The paper presents a learning method to learn the mapping from an input space to an action space, which is particularly suitable when the action is an optimal decision with respect to a certain unknown cost function. I use an inverse optimization approach to retrieve the cost function by introducing a new loss function and a new hypothesis class of mappings. A tractable convex reformulation of the learning problem is also presented. The method is effective for learning input-action mapping in continuous input-action space with input-output constraints, typically present in control systems. The learning approach can be effectively transformed to learn a Model Predictive Control (MPC) behaviour and a case study to mimic an MPC is presented, which is a rather computationally heavy control strategy.

Spherical Robot for Search and Reconnaissance

The aim was to develop a remotely operated robot equipped with vision sensor for search and reconnaissance applications. A spherical robot comprises of a ball shaped outer shell, which holds the entire mechanism, actuators, and electronics. It has several advantages over wheeled and legged robots. Owing to its spherical shape, it cannot overturn as opposed to wheeled and legged robots. It is maneuverable in any direction and can easily recover from collisions and obstacles. The spherical shape also eases the deployment strategy. It can be thrown, kicked or dropped at the site of operation with no risk of overturning or extruding actuator and electronics damage. This project uses the spherical robot developed in embedded control laboratory of systems and control engineering group, IIT Bombay. The design of this spherical robot is gearless and scalable in size. Hence, this project investigates its applications in various environments. The manually operated spherical robot would provide visual feedback and other application specific information to the operator for reconnaissance and search. Please find the git repo here