Research Projects

Robotics has numerous applications in various fields, including engineering, aerospace, modern industries, medicine, and disaster management. Wheeled robots are prevalent in robotics due to their low energy consumption, high speed, and small control effort required. Motion planning in wheeled robotics systems requires efficient motion control algorithms, which are essential in disaster response due to difficult geographical conditions, congested terrains, and inclement weather. Traditional strategies like PID control, optimal control, and adaptive control have drawbacks such as convergence, stability, lack of efficient tracking, and time consumption. A combined approach called adaptive optimal control (AOC) technique has provided satisfactory results in terms of desired optimal control action in the absence of system dynamics while overcoming common issues with previous algorithms. The current proposal aims to develop efficient machine learning-based AOC algorithms to regulate wheeled robots used in disaster response scenarios. Reinforcement learning (RL) is a popular domain of machine learning for solving AOC problems, but most existing AOC schemes for continuous-time (CT) systems suffer from issues like the requirement of complete/partial knowledge of system dynamics, restrictive conditions of persistence of excitation, and memory deficiency. The project proposal aims to develop machine learning/RL-based AOC algorithms for wheeled robot locomotion, addressing issues like complete/partial knowledge of system dynamics, restrictive conditions of persistence of excitation, and memory deficiency. The new algorithms will be implemented on a robotic test platform and analyzed for significantly efficient wheeled robotic locomotion in disaster response activities.