SpikingSoft: A Spiking Neuron Controller for Bio-inspired Locomotion with Soft Snake Robots

TL;DR

This work introduces SpikingSoft, a model-free gait controller for soft snake robots that uses a novel Double Threshold Spiking (DTS) neuron to harness the robot's intrinsic oscillations. By adjusting two thresholds, $u_n$ and $u_p$, the DTS neuron generates versatile spike patterns that drive segment torques, enabling smooth, reactive locomotion when integrated with reinforcement learning. The approach demonstrates substantial performance gains over vanilla RL and CPG controllers, including higher success rates, faster target reaching, and improved energy efficiency, and proves scalable to longer snake configurations. The findings suggest that bio-inspired, threshold-based spiking neurons can effectively control high-DOF soft robots without explicit gait models, with potential for real-world applications in constrained environments.

Abstract

Inspired by the dynamic coupling of moto-neurons and physical elasticity in animals, this work explores the possibility of generating locomotion gaits by utilizing physical oscillations in a soft snake by means of a low-level spiking neural mechanism. To achieve this goal, we introduce the Double Threshold Spiking neuron model with adjustable thresholds to generate varied output patterns. This neuron model can excite the natural dynamics of soft robotic snakes, and it enables distinct movements, such as turning or moving forward, by simply altering the neural thresholds. Finally, we demonstrate that our approach, termed SpikingSoft, naturally pairs and integrates with reinforcement learning. The high-level agent only needs to adjust the two thresholds to generate complex movement patterns, thus strongly simplifying the learning of reactive locomotion. Simulation results demonstrate that the proposed architecture significantly enhances the performance of the soft snake robot, enabling it to achieve target objectives with a 21.6% increase in success rate, a 29% reduction in time to reach the target, and smoother movements compared to the vanilla reinforcement learning controllers or Central Pattern Generator controller acting in torque space.

Figures (8)

• Figure 1: The pipeline of the proposed biologically plausible control architecture. Left: The SpikingSoft controller structure. The oscillatory behavior $d$ is the input of the proposed spiking neuron. $C_{\mathrm{t}}$ and $C_{\mathrm{q}}$ are hyperparameters to adjust the scale. $o$ is the output spikes. Right: The reinforcement learning framework of a 3-segment 3-node snake with SpikingSoft controllers. $\mu^{\mathrm{i}}$ and $\sigma^{\mathrm{i}}$ are taken from the RL actions, composing the double thresholds $u_{\mathrm{n}}^{\mathrm{i}}$ and $u_{\mathrm{p}}^{\mathrm{i}}$ for the controller in the $i^{\mathrm{th}}$ segment.
• Figure 2: The phase space limit circle of the mass-spring-damper system. a: The mass-spring-damper system. Others: Three oscillation patterns controlled by a DTS neuron. b:$u_{\mathrm{n}}=-0.1, u_{\mathrm{p}}=-0.025$. c:$u_{\mathrm{n}}=-0.0025, u_{\mathrm{p}}=0.0025$.
• Figure 3: Example of DTS neuron dynamics. In this example, the hyperparameter $C_{\mathrm{q}}$ is 10 and $\tau$ is 0.1. The sampling interval $\mathrm{d}t$ is 0.001s and the simulation duration is 2s. The input $q$ is a sine function shown in orange. The weighted input $\frac{\mathrm{d}t}{\tau} C_{\mathrm{q}} q^\mathrm{k+1}$ is described in red. The membrane potential $u$ is black. Both positive and negative spikes are shown as scatter lines. This example covers three basic situations of a DTS neuron.
• Figure 4: Example of the DTS neuron (left) and snake segment states with the SpikingSoft controller (right). $u_{\mathrm{p}}$ is 0.2 and $u_{\mathrm{n}}$ is -0.2. At time step 0, the snake segment has an initial positive deformation.
• Figure 5: Left: Example of the DTS neuron state when $u_{\mathrm{n}}$ and $u_{\mathrm{p}}$ are of the same sign. The figure shows three cases: 1) Green represents $u_{\mathrm{n}}$ equals 0.05 and $u_{\mathrm{p}}$ equals 0.075; 2) Blue means $u_{\mathrm{n}}$ equals 0.05 and $u_{\mathrm{p}}$ equals 0.15; 3) Red represents $u_{\mathrm{n}}$ equals 0.1 and $u_{\mathrm{p}}$ equals 0.15. Right: Three examples of the segment state with the SpikingSoft controller when $u_{\mathrm{n}}$ and $u_{\mathrm{p}}$ are of the same sign, which corresponds to left figure.