Multi-Modal LSTM-Based Tracking Control for a Soft Robotic Manipulator with Hysteresis

Role

MSc Researcher, then Guest Researcher (remote), Hamlyn Centre for Robotic Surgery, Imperial College London. December 2024 to July 2026.

Problem

Hysteresis in pneumatically driven soft surgical manipulators limits how accurately a model-based controller can track a reference.

Approach

A hybrid LSTM/physics-based model of the hysteresis, built into a tracking controller. The control variables are the pressure differentials between chambers; the controller ran in real time in MATLAB and was validated hardware-in-the-loop.

Result

Tracking error reduced by 45% against an optimally tuned, purely model-based controller.

Paper

B. D. Bassett, A. Dias, J. Shi, F. Shakib, E. Franco, K. Chen, “Multi-Modal LSTM-Based Tracking Control for a Soft Robotic Manipulator with Hysteresis,” Proc. 2026 European Control Conference (ECC), Reykjavík, Iceland, 2026, pp. 182–187.

Write-up in progress

Figures and a detailed account are coming. In the meantime, I'm happy to talk through the work.