Research Analyzer
← Back ICRA 2026

State Estimation for Compliant and Morphologically Adaptive Robots

Valentin Yuryev, Max Polzin, Josie Hughes

PDF

AI summary

Key figure (auto-extracted from paper)
A neural network-based estimator accurately tracks both rigid-body states and complex morphing shapes, enabling robust autonomous control for compliant robots even with sensor failures.
state estimation compliant robots morphological adaptation neural networks soft robotics autonomous control

Problem

State estimation for compliant, morphologically adaptive robots is challenging due to the lack of rigid-body assumptions, continuous kinematic changes, and unreliable sensor data, which hinders reliable autonomous control in extreme terrains.

Approach

The authors introduce a history-based recurrent neural network that estimates a compliance-centric reference frame, outer shell shape, and velocities using only IMU and motor data, while dynamically correcting untrustworthy sensor inputs.

Key results

  • Predicts shape measurements within 4.2% of robot size
  • Estimates linear and angular velocities within 6.3% and 2.4% of top speeds
  • Achieves orientation accuracy within 1.5 degrees
  • Enables 300% increase in travel range during motor malfunctions

Why it matters

This approach enables reliable autonomy for next-generation compliant robots navigating unpredictable outdoor environments where traditional rigid-body estimators fail.

Abstract

Locomotion robots with active or passive com- pliance can show robustness to uncertain scenarios, which can be promising for agricultural, research and environmen- tal industries. However, state estimation for these robots is challenging due to the lack of rigid-body assumptions and kinematic changes from morphing. We propose a method to estimate typical rigid-body states alongside compliance-related states, such as soft robot shape in different morphologies and locomotion modes. Our neural network-based state esti- mator uses a history of states and a mechanism to directly influence unreliable sensors. We test our framework on the GOAT platform, a robot capable of passive compliance and active morphing for extreme outdoor terrain. The network is trained on motion capture data in a novel compliance-centric frame that accounts for morphing-related states. Our method predicts shape-related measurements within 4.2% of the robot’s size, velocities within 6.3% and 2.4% of the top linear and angular speeds, respectively, and orientation within 1.5◦. We also demonstrate a 300% increase in travel range during a motor malfunction when using our estimator for closed-loop autonomous outdoor operation.

Index terms

Field Robots Sensor Fusion Deep Learning Methods

Related papers