Adaptive Control of Soft Continuum Robots

Robust adaptive dynamic sliding mode control for task-space tracking of soft continuum manipulators under payload uncertainty.

Outcome

I designed and validated an adaptive dynamic sliding mode control framework for a two-segment soft continuum manipulator, targeting accurate task-space tracking under payload uncertainty. On hardware, the controller achieved about 38% better tracking accuracy than inverse-dynamics baselines across trajectory and payload variations.

Problem

Soft continuum manipulators are compliant and highly nonlinear. In practice, their effective dynamics shift with payload and operating conditions, so model mismatch can quickly degrade tracking performance. The goal in this project was to keep tracking accurate on physical hardware while remaining robust to uncertainties and unknown disturbances.

System

  • Robot platform: two-segment pneumatic soft continuum arm with a soft gripper and motion-capture feedback
  • Modeling: Euler-Lagrange dynamic formulation with centroid-based mass modeling and linear-in-parameters structure
  • Control law: adaptive dynamic sliding mode control with online parameter adaptation and disturbance-bound compensation
  • Benchmarks: inverse-dynamics control baseline evaluated under matched trajectories and payload settings
  • Implementation: real-time task-space closed-loop experiments on physical hardware

Contribution

  • Developed the model-based adaptive control architecture for the continuum arm
  • Derived and implemented adaptation laws and robust sliding-mode terms for disturbance handling
  • Designed and ran hardware experiments across multiple trajectories, speeds, and payload conditions
  • Quantitatively benchmarked adaptive control against inverse-dynamics baselines

Technical Stack

  • Nonlinear robot dynamics (Euler-Lagrange)
  • Adaptive dynamic sliding mode control
  • Task-space control and trajectory tracking
  • Motion-capture-based feedback integration
  • C++ real-time implementation and experimental evaluation

Key Results

  • About 38% tracking-accuracy improvement versus inverse-dynamics control under payload variation
  • Robust tracking maintained across both circular and star-shaped task-space trajectories
  • Evaluated under multiple payload conditions (including 12 g and 25 g) and both slow/fast timing profiles
  • Dynamic-term update with the proposed Lagrangian formulation reported about one order of magnitude faster than the augmented rigid-body benchmark implementation
  • Published at IEEE ICRA (2022)

Media

Video demo: adaptive dynamic sliding mode control on a soft continuum manipulator under trajectory and payload variation.

Impact and Future Direction

This work shows that model-based adaptive control can deliver reliable task-space performance on soft continuum hardware, even when payload conditions change. The same control structure can be extended to multi-segment continuum systems and manipulation tasks where dynamic uncertainty is unavoidable.

Skills

soft continuum robotics adaptive control sliding mode control nonlinear dynamics task-space tracking hardware validation