These assignments constitute an integral component of the Embodied AI course. They are specifically designed to ensure that students develop a thorough understanding of the theoretical principles and methodological foundations of Embodied Artificial Intelligence. Furthermore, the assignments aim to enable students to effectively apply the knowledge and skills they have acquired to the implementation of computational tasks and practical real-world applications, thereby reinforcing the connection between theory and practice. To support this objective, the assignments will be conducted using the CoppeliaSim simulation environment, with control algorithms implemented through Lua scripting, allowing students to design, test, and evaluate embodied controllers in a realistic and interactive simulated setting.
Task 1: Unsupervised learning (Input correlation (ICO) learning) for prediction "seeing future" and anticipation "acting on it": An application in goal-directed mobile robot navigation. --> Tasks: student_ICO_2_wheeled_mobile_robot.ttt
Task 2: Supervised learning (Error-backpropagation learning) for nonlinear function approximation (XOR): Applied to tower defense control logic. --> Tasks: student_XOR_tower.ttt
Task 3: Reinforcement learning (Q/SARSA learning) for state-action mapping: An application in autonomous obstacle avoidance and decision-making. --> Tasks: student_Q_learning_2_wheeled_mobile_robot.ttt