Low-Load endoluminal navigation with a magnetically actuated medical soft microrobot

Minimally invasive endoluminal interventions increasingly rely on magnetic actuation to navigate narrow lumens while limiting wall loads. Here we present a contact-aware control framework for steering a deformable, silicone-based soft microrobot with embedded magnetic particles using an externally positioned permanent magnet. We develop a dynamic model capturing viscous drag, nonlinear frictional loads, and viscoelastic wall contact, and implement a closed-loop architecture that combines vision-based state estimation with model-based force inference while optimizing magnet orientation to regulate the force vector and normal reaction. Performance is evaluated in simulation and on a benchtop testbed across three control modes. In the nominal-case benchmark, force-plus-angle control reduced the root-mean-square tracking error from 4.8 to 2.1 mm (−56%), decreased peak tracking error from 14.6 to 8.0 mm (−45%), lowered the integrated performance index from 4.4 × 10−¹⁰ to 1.6 × 10−¹⁰ (−64%), and attenuated peak normal reaction force from 2.0 × 10−6 to 0.8 × 10−6 N (−60%) compared with operation without force regulation. To assess robustness, we further performed a simulation-based Monte Carlo analysis (n = 500 trials per mode) under parametric uncertainty and measurement noise, confirming that the contact-aware modes preserve their performance advantage; non-parametric tests indicated statistically significant inter-mode differences with moderate-to-large effect sizes. A trade-off analysis in the {tracking error, peak normal load} plane showed that, in the explored regime, improved tracking does not inherently require higher peak contact forces. Finally, a first-order shear-thinning surrogate suggested low sensitivity of the relative conclusions to moderate non-Newtonian effects. Overall, the results identify force-aware magnet orientation as a safety-relevant control degree of freedom for endoluminal navigation and provide a transferable control methodology for future magnetic microrobotic platforms.