State Estimation for Compliant and Morphologically Adaptive Robots

TL;DR

Conventional state estimation assumes rigid bodies, which is limiting for compliant, morphologically adaptive robots. The paper proposes a learning-based estimator using a history-based belief encoder within a compliance-centric frame to jointly infer rigid-body states and morphology-specific states (e.g., outer-shell shape) from onboard sensors. Trained with motion capture ground truth on the GOAT platform, the method achieves accurate shape, velocity, and orientation estimates and improves closed-loop control, enabling substantial outdoor operation gains under fault conditions. This work demonstrates morphology-aware sensing as a pathway to robust, multi-modal locomotion in challenging environments.

Abstract

Locomotion robots with active or passive compliance can show robustness to uncertain scenarios, which can be promising for agricultural, research and environmental 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 estimator 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 degrees. We also demonstrate a 300% increase in travel range during a motor malfunction when using our estimator for closed-loop autonomous outdoor operation.

Figures (10)

• Figure 1: Overview of our state estimation framework for compliant and morphologically adaptive platforms. Left: placement of motors, sensors and motion capture for data collection. Right: reconstruction of compliance-centric frame, shape and width.
• Figure 2: Belief encoder architecture. Hidden states are passed through onto the next iteration. Inputs are fully fed into the RNN while a subset of the inputs that are considered untrustworthy are further adjusted by the network.
• Figure 3: Overlay of the reconstructed frame shape over the actual robot image for scenarios in Table \ref{['tab:recon_rmse']}. Note that tendon lengths estimated specified in case (D) and (E) are the same. Yet the framwork is capable of reconstruction the shape.
• Figure 4: Time series of robot morphing from rover (scenario B) to sphere (scenario C) formation.
• Figure 5: (Top) Time lapse images of the GOAT robot rolling down a hill and (Bottom) the estimated gravity vector time series.