Shape-Aware Whole-Body Control for Continuum Robots with Application in Endoluminal Surgical Robotics

TL;DR

This work addresses safe, accurate endoluminal navigation with tendon-driven continuum robots by combining a physics-informed Cosserat backbone model with learned residuals via an Augmented Neural ODE, enabling full-body shape estimation. A sampling-based MPPI controller jointly optimizes tip tracking, backbone conformance, and obstacle avoidance under actuation constraints, while a Task Manager allows real-time adjustment of objectives during tele-operation. The approach achieves millimeter-level tracking and robust shape coordination in simulations, and demonstrates reduced wall contacts and improved target access in a bronchoscopy phantom compared to joystick-only control and baselines. The results suggest a practical, adaptable framework for shape-aware continuum robot control in confined and safety-critical environments, with potential extensions to dynamic anatomies and intent-aware planning.

Abstract

This paper presents a shape-aware whole-body control framework for tendon-driven continuum robots with direct application to endoluminal surgical navigation. Endoluminal procedures, such as bronchoscopy, demand precise and safe navigation through tortuous, patient-specific anatomy where conventional tip-only control often leads to wall contact, tissue trauma, or failure to reach distal targets. To address these challenges, our approach combines a physics-informed backbone model with residual learning through an Augmented Neural ODE, enabling accurate shape estimation and efficient Jacobian computation. A sampling-based Model Predictive Path Integral (MPPI) controller leverages this representation to jointly optimize tip tracking, backbone conformance, and obstacle avoidance under actuation constraints. A task manager further enhances adaptability by allowing real-time adjustment of objectives, such as wall clearance or direct advancement, during tele-operation. Extensive simulation studies demonstrate millimeter-level accuracy across diverse scenarios, including trajectory tracking, dynamic obstacle avoidance, and shape-constrained reaching. Real-robot experiments on a bronchoscopy phantom validate the framework, showing improved lumen-following accuracy, reduced wall contacts, and enhanced adaptability compared to joystick-only navigation and existing baselines. These results highlight the potential of the proposed framework to increase safety, reliability, and operator efficiency in minimally invasive endoluminal surgery, with broader applicability to other confined and safety-critical environments.

Figures (9)

• Figure 1: Overall architecture of the proposed framework, consisting of three main modules. The backbone modeling combines physics-based Cosserat modeling with neural residuals to provide accurate backbone reconstruction. The Whole-Body Controller employs MPPI to jointly optimize tip tracking, shape regulation, and obstacle avoidance under actuation constraints. The Task Manager interfaces with the operator, enabling online adjustment of control objectives such as clearance, compliance, or direct advancement.
• Figure 2: Representative results for the trajectory tracking scenario: red markers show the current tip position, green dashed lines are the reference trajectories, and solid lines indicate the robot's body shape: (a) circular, (b) curvy- edge circular, (c) butterfly, (d) elliptical, and (e) helical trajectories.
• Figure 3: A representative simulation result of Dynamic obstacle avoidance. The green curve represents the reference trajectory, and the red marker denotes the current tip position. Blue and magenta markers indicate obstacles. The solid curve illustrates the three connected body segments of the robot, highlighting how the framework coordinates whole-body motion to remain close to the desired trajectory while maintaining safe clearance from obstacles.
• Figure 4: Performance of the dynamic obstacle avoidance task. The controller maintains millimeter-level tracking accuracy with error spikes during avoidance maneuvers. The control effort gradually increases as the robot reshapes its body to ensure clearance, while the minimum distance to obstacles consistently remains above the 0.02 m safety threshold, demonstrating successful collision avoidance.
• Figure 5: Simulation results of the Shape-constrained reaching task. The transparent green lines represent the reference body shapes, while the red markers indicate the target tip positions. The robot adapts its configuration to simultaneously match the desired tip location and overall shape.