New catheter control model slashes force errors by 90% with hard constraints

A new preprint from Yongyan Cao tackles a critical safety challenge in robotic catheterization: keeping the tip’s contact force below clinically safe limits while maintaining precise tracking. The paper formulates catheter–tissue interaction as a scalar, configuration‑invariant dynamics model. A partial‑physics feedforward cancels nominal bending, leaving a linear interaction model with varying input gain. A predictive optimizer then regulates this state subject to hard constraints on contact force, tendon force, and curvature. An augmented Kalman filter lumps friction, contact, and modeling errors into a single disturbance state, enabling offset‑free motion in free space without sacrificing constraint enforcement.

In MuJoCo simulations of an eight‑link tendon‑driven catheter, the disturbance‑augmented filter reduced free‑space approach error by 90%. Under stiff tissue contact without constraints, the controller drove force to 0.60 N—above the 0.5 N bound. With explicit predictive constraints, force held at 0.47 N at identical tracking accuracy. The bound also remained valid under 0.5 mm, 1.2 Hz simulated cardiac motion. These results show that motion regulation and contact safety are coupled objectives, resolvable only through constrained predictive control. Hardware validation is planned as future work.