To the best of our knowledge, this framework is the first bio-inspired neural controller deployed on a distributed embedded system.
With relatively fewer neurons than more complex life forms, insects are still capable of producing astonishing locomotive behaviors, such as traversing diverse habitats and making rapid gait adaptations after extreme injury or autotomy. Biologists attribute this to a chain of segmental neuron clusters (ganglia) within insect nervous systems, which act as distributed, self-organizing sensorimotor control units. Inspired by the neural structure of the Carausius morosus, the common stick insect, this research introduces the Distributed Neural Locomotion Controller (D-NLC), a modular control framework utilizing local proprioceptive feedback to modulate joint-level Central Pattern Generator (CPG) signals to produce emergent locomotive behaviors.
We implemented this framework using a modular legged robot with distributed joint-level embedded computing units and assessed its performance and behavior under various experimental settings. Based on real-world experiments, we observe an overall 31.3% average increase in curvilinear motion performance under external (terrain) and internal (amputation) perturbation compared to a centralized predefined gait controller. This difference is statistically significant (P<<0.05) for larger perturbations but not for single-leg amputations. Experiments with perturbation-induced leg stance duration and leg-phase-difference analysis further validated our hypothesis regarding D-NLC's role in the robust perceptive locomotion and self-emergent gait adaptation against complex unforeseen perturbations. This proposed control framework does not require any numerical optimization or weight training processes, which are time-consuming and computationally expensive.
Inspired by a stick insect, D-NLC coordination rules mimic biological proprioceptive feedback, dynamically adjusting gait based on sensory input from interleg, intraleg signals. EigenBot achieves adaptive locomotion without predefined patterns or cost functions as seen in Reinforcement Learning or Optimal Control.
EigenBot achieves stable flat walking using D-NLC, leveraging proprioceptive feedback and joint-level CPG modulation for smooth, adaptive movement. It maintains a consistent tripod-like gait, optimizing efficiency without predefined motion patterns. This decentralized control enables robust and responsive locomotion, closely resembling biological walking.
EigenBot traverses uneven terrain by dynamically adjusting stance duration and gait coordination, allowing for stable locomotion over slopes, obstacles, and unpredictable surfaces. Longer stance phases improve balance, while emergent interlimb adaptations prevent instability. This self-organized response ensures continuous movement without external intervention or reprogramming.
D-NLC enables self-organized gait recovery by dynamically adjusting interlimb coordination after limb loss. Redistributing proprioceptive influences, it extends stance phases and shifts phase relationships for balance. This often results in a trotting-like gait, optimizing stability and forward motion. The adaptive response ensures smooth locomotion even with significant limb perturbations.
Inspired by insect locomotion, D-NLC facilitates curve walking by integrating a modular neuron into the ThC joint network. This neuron modifies the inner middle leg’s movement, allowing it to switch its heading. Upon activation, it reverses leg direction or adjusts protractor-retractor range for precise turns. Tested in simulation, this adaptation ensures agile and controllable navigation in varied environments.
EigenBot’s modular design allows flexible reconfiguration, minimizing cost and complexity while enabling specialized tasks. Developed by the CMU Biorobotics Lab, its EigenBus communication protocol ensures real-time distributed control. Each module features onboard sensors, a microcontroller, and a tool-less octo-indent interface for rapid deployment. This modularity makes it ideal for distributed embedded control.
Slide the scrubber to view CPG activity and neuron activations during gait cycles of leg R1.
@article{zhang2025dnlc,
author = {Zhang, Guo, Kou, Shikhare, Choset, Li},
title = {Bio-inspired Distributed Neural Locomotion Controller (D-NLC) for Robust Locomotion and Emergent Behaviors},
journal = {ICRA},
year = {2025},
}