RL with QUANSER Qube-Servo
Implementing from scratch Q-learning, SARSA, DQN, REINFORCE, A2C, DDPG, TD3 and SAC without any library. MATLAB, CPU-based training. Algorithms deployed in the QUANSER Qube-Servo.
Jul 1, 2025
Trajectory Optimization for dynamic feasibility of LLM-derived human-like movements on a NAO robot
Transforming kinematic trajectories obtained by an LLM-like Generative Masked Modeling (MoMask) into dynamically feasible trajectories, which comply with dynamic restrictions related to friction and contact for legged locomotion. Used trajectory optimization in Ipopt C++.
Apr 1, 2025
Nonlinear Control Labs
Implementing nonlinear controllers in hardware using C++ code in a 4dof Dynamixel OpenManipulator-X robot. Controllers: Feedback Linearization, Trajectory Optimization, Sliding Mode Control and Adaptive Control.
Feb 1, 2025