Amir Suhail Salim

Thesis: Learning-Based End-Effector Trajectory Tracking for Underwater Vehicle-Manipulator Systems (BlueROV2 Heavy + Reach Alpha 5). • Initialising neural control from a real-system-validated TPIK + PID-C expert via behavior cloning + DAgger • Fine-tuning with TD3+BC in simulation (privileged critic; domain randomization of payload, thruster lag, sensing delay, noise) • ROS 2 policy node with a deterministic safety layer outside the network for staged real-system validation

• Used a provided NMPC expert (CasADi/IPOPT) to generate 10,000 demonstrations from 125 closed-loop episodes (9–15.5 bar) • Trained a compact 56-parameter behavior-cloning policy: 0.002 held-out valve-command RMSE, NMPC-comparable pressure tracking • Cut mean controller computation time from 60.6 ms to 0.32 ms (~190× faster)

• Developed Pytest-based automated test frameworks, test specifications and acceptance criteria for engineering software modules • Automated logging and report generation for engineering software modules

In-plant training: CNC machining and G-code programming.

Focus: control systems, autonomous robotics, MPC, reinforcement learning, energy systems. Master's thesis (MIRMI): Learning-Based End-Effector Trajectory Tracking for Underwater Vehicle-Manipulator Systems. Expected completion: January 2027.

Bachelor's thesis: Battery Thermal Management Systems for EVs (CFD simulations in COMSOL Multiphysics). KI Racing (Baja SAE) team member, Dec 2018 – Mar 2020: front-wheel assembly design and manufacturing.