[Paper Review] Multi-tap Resistive Sensing and FEM Modeling enables Shape and Force Estimation in Soft Robots

Sumei Tian, Barnabas Gavin Cangan+4 more|arXiv (Cornell University)|Nov 24, 2023

Soft Robotics and Applications1 citations

TL;DR

This paper proposes a multi-tap resistive sensing approach combined with finite element modeling (FEM) to enable high-resolution proprioception in soft robots using off-the-shelf sensors. By embedding a multi-tapped flex sensor into a soft robotic structure and using an FEM model that accounts for fluidic actuation, the system estimates shape with ~3% average relative error and external force with 11% relative error over a 0–5 N range, enabling closed-loop manipulation without external tracking.

ABSTRACT

We address the challenge of reliable and accurate proprioception in soft robots, specifically those with tight packaging constraints and relying only on internally embedded sensors. While various sensing approaches with single sensors have been tried, often with a constant curvature assumption, we look into sensing local deformations at multiple locations of the sensor. In our approach, we multi-tap an off-the-shelf resistive sensor by creating multiple electrical connections onto the resistive layer of the sensor and we insert the sensor into a soft body. This modification allows us to measure changes in resistance at multiple segments throughout the length of the sensor, providing improved resolution of local deformations in the soft body. These measurements inform a model based on a finite element method (FEM) that estimates the shape of the soft body and the magnitude of an external force acting at a known arbitrary location. Our model-based approach estimates soft body deformation with approximately 3% average relative error while taking into account internal fluidic actuation. Our estimate of external force disturbance has an 11% relative error within a range of 0 to 5 N. The combined sensing and modeling approach can be integrated, for instance, into soft manipulation platforms to enable features such as identifying the shape and material properties of an object being grasped. Such manipulators can benefit from the inherent softness and compliance while being fully proprioceptive, relying only on embedded sensing and not on external systems such as motion capture. Such proprioception is essential for the deployment of soft robots in real-world scenarios.

Motivation & Objective

• To address the challenge of reliable proprioception in soft robots with tight packaging and only internal sensors.
• To improve shape and force estimation accuracy beyond single-sensor, constant-curvature assumptions.
• To enable high-resolution local deformation sensing using a modified, multi-tapped resistive sensor.
• To develop a model-based approach that integrates actuation inputs and sensor data for accurate state estimation.
• To demonstrate real-time, embedded proprioception for soft grippers in unstructured, human-centric environments.

Proposed method

• Modifying an off-the-shelf resistive flex sensor by creating multiple electrical taps along its conductive layer to enable multi-segment resistance measurement.
• Using a neural network to decode the nonlinear, coupled resistance signals from the multi-tap sensor into local shape descriptors (position and angular displacement).
• Developing a finite element method (FEM) model in SOFA that simulates the mechanical behavior of the soft robot, including fluidic actuation and embedded sensor geometry.
• Integrating the multi-tap sensor readings as boundary conditions in the FEM model to estimate full-body shape and external forces.
• Calibrating the FEM model using experimental data from flexible strips and soft pressurized fingers to improve accuracy.
• Validating the approach using both vision-based ground truth and force-torque sensors for shape and force estimation.

Experimental results

Research questions

• RQ1 Can multi-tap resistive sensing improve local deformation resolution compared to single-point sensing in soft robots?
• RQ2 Can an FEM-based model accurately reconstruct the full shape of a soft robot using only embedded sensor data and actuation inputs?
• RQ3 To what extent can the system estimate external forces acting at arbitrary locations with minimal error?
• RQ4 How does the model perform under varying actuation pressures and external loads?
• RQ5 Can the approach generalize across different object shapes and contact points without retraining?

Key findings

• The system achieves an average relative shape estimation error of approximately 3% across the full length of the soft robot.
• External force estimation has a relative error of 11% over a 0–5 N range, with consistent performance across different contact locations.
• The multi-tapped sensor enables full shape reconstruction beyond the active length of a single sensor, unlike single-tap methods.
• The FEM-based model generalizes well to new object shapes and contact points without requiring retraining or new data collection.
• The model runs in real-time at 30 Hz for flexible strips and 5 Hz for the more complex soft finger, demonstrating feasibility for embedded control.
• The approach outperforms data-driven methods in range and relative error, with a 5 N range compared to 0.5 N in prior work, despite using a model-based approach.