Magnetic Nanorobot Path Control

OpenCVFeedback ControlTrajectory Tracking

Developed a closed loop trajectory control system for a magnetically actuated helical nanorobot designed for targeted drug delivery. Overhead vision estimates the robot's pose, while a PD controller converts path tracking error into magnetic torque commands, driving corkscrew propulsion along 2D trajectories.

System

01Problem

Develop a closed loop control scheme capable of guiding a magnetically actuated helical nanorobot along 2D trajectories for targeted drug delivery applications.

Existing magnetic nanorobot approaches often rely on manual steering, open loop fields, or simplified path planning, limiting their ability to continuously correct deviations from an arbitrary trajectory.
Small position and heading errors must be corrected continuously because the robot's direction of propulsion is coupled directly to the orientation of the applied rotating magnetic field.
A useful controller therefore needs to connect visual localization, trajectory tracking, magnetic actuation, and low Reynolds number fluid dynamics in one feedback loop.

02Approach

Built a vision based trajectory controller that converts robot pose error into the magnetic field commands required to steer a helical microrobot.

Used overhead camera feedback to recover the robot's position and heading, then computed cross track and heading error relative to the closest point and tangent of the desired trajectory.
Applied look ahead trajectory geometry and PD control to generate steering and torque commands while maintaining forward motion and stable path convergence, including at low speeds.
Derived the full actuation transform from commanded torque → magnetic field → coil current → voltage, connecting the controller directly to the Helmholtz-coil actuation architecture.

03Result

Demonstrated a complete closed loop simulation in which the nanorobot autonomously followed curved trajectories using feedback driven magnetic steering.

Modeled the robot's rotational and translational response using low Reynolds number resistance dynamics, allowing magnetic torque to propagate into angular velocity, forward speed, and planar motion.
The simulated robot closely followed the commanded curved path with minimal steady state error while continuously correcting its heading from the measured trajectory deviation.
The final framework connected trajectory generation, PD control, and robot dynamics into one feedback architecture for future hardware implementation.