Intelligent Collective Escape of Swarm Robots Based on a Novel Fish-inspired Self-adaptive Approach with Neurodynamic Models

TL;DR

This work tackles reliable collective escape for swarm robots under dynamic threats with limited sensing and no inter-robot communication. It introduces a fish-inspired framework that combines a bio-inspired neural network (BINN) to generate attractive and repulsive virtual forces with a neurodynamics-based self-adaptive mechanism to cope with changing environments. Key contributions include three behavioral modes (Align, Escape, Follow), a hierarchical organization to manage large swarms, and a velocity-mapped BINN-based motion strategy that remains collision-free. Through simulations and real-robot experiments, the approach achieves higher success rates, lower escape time, and reduced energy consumption compared with state-of-the-art methods, demonstrating strong practical potential for rescue and surveillance swarms in complex settings.

Abstract

Fish schools present high-efficiency group behaviors through simple individual interactions to collective migration and dynamic escape from the predator. The school behavior of fish is usually a good inspiration to design control architecture for swarm robots. In this paper, a novel fish-inspired self-adaptive approach is proposed for collective escape for the swarm robots. In addition, a bio-inspired neural network (BINN) is introduced to generate collision-free escape robot trajectories through the combination of attractive and repulsive forces. Furthermore, to cope with dynamic environments, a neurodynamics-based self-adaptive mechanism is proposed to improve the self-adaptive performance of the swarm robots in the changing environment. Similar to fish escape maneuvers, simulation and experimental results show that the swarm robots are capable of collectively leaving away from the threats. Several comparison studies demonstrated that the proposed approach can significantly improve the effectiveness and efficiency of system performance, and the flexibility and robustness in complex environments.

Figures (11)

• Figure 1: The flowchart of the model transmission. Mode 1: the swarm robots detect the environment, including information on the threats, obstacles, and other robots. Mode 2: if one robot detects the threat, this robot records the position of the threat and generates the escape trajectory. Mode 3: the resultant force combines the attractive and repulsive forces, where two self-adaptive weights are used to adjust the influence of the attractive and repulsive forces.
• Figure 2: The illustration of fish-inspired collective organization. The robots cannot share information with neighbors, only the first robot to detect the threat is aware of the threat position.
• Figure 3: Examples of the bio-inspired neural network. (a) structure of the neural network with only local connections; (b) the dynamic landscape of neural activity.
• Figure 4: Robots escape in a static environment. (a) the initial position of robots; (b) robots escape at time $8s$; (c) robots escape at 16s; (d) robots escape at $43s$; (e) robots escape at $52s$.
• Figure 5: Robots escape in a dynamic environment. (a) the initial position of robots and moving obstacles; (b) robots escape at $7s$; (c) robots escape at $14s$; (d) robots escape at $36s$; (e) robots escape at $45s$.