Spiking Neural Networks as a Controller for Emergent Swarm Agents

TL;DR

The feasibility of training spiking neural networks to find those local interaction rules that result in particular emergent behaviors in swarms of robots with only a binary sensor and a simple but hand-picked controller structure is investigated.

Abstract

Drones which can swarm and loiter in a certain area cost hundreds of dollars, but mosquitos can do the same and are essentially worthless. To control swarms of low-cost robots, researchers may end up spending countless hours brainstorming robot configurations and policies to ``organically" create behaviors which do not need expensive sensors and perception. Existing research explores the possible emergent behaviors in swarms of robots with only a binary sensor and a simple but hand-picked controller structure. Even agents in this highly limited sensing, actuation, and computational capability class can exhibit relatively complex global behaviors such as aggregation, milling, and dispersal, but finding the local interaction rules that enable more collective behaviors remains a significant challenge. This paper investigates the feasibility of training spiking neural networks to find those local interaction rules that result in particular emergent behaviors. In this paper, we focus on simulating a specific milling behavior already known to be producible using very simple binary sensing and acting agents. To do this, we use evolutionary algorithms to evolve not only the parameters (the weights, biases, and delays) of a spiking neural network, but also its structure. To create a baseline, we also show an evolutionary search strategy over the parameters for the incumbent hand-picked binary controller structure. Our simulations show that spiking neural networks can be evolved in binary sensing agents to form a mill.

Figures (7)

• Figure 1: Diagram explaining sensor variables and sensor placement
• Figure 2: Diagram of encoding process for a single robot. Top: Sensor not activated; robot does not "see" another agent. Bottom: Sensor activated, robot "sees" other agent in its FOV.
• Figure 3: Top: Population Best Fitness per Epoch. For the symbolic 4-parameter controller optimized with CMA-ES, a single run was performed, and the search was terminated at the 263rd epoch. For the SNNs evolved with EONS, 5 runs with different evolutionary seeds were performed. The shaded silhouette bounds the min/max fitness achieved for a particular epoch across the EONS runs. The EONS run which achieved the highest final fitness is highlighted. The annotations indicate when the highest fitness score was first achieved in the best run. Bottom: Focused view of Top plot.
• Figure 4: Top: Distribution of fitness scores in the population after each epoch. Each dot represents the fitness of a single simulation in RSSim and is transparent (opacity = 0.2). For EONS, only data from the best run is shown. Bottom: Number of neurons and synapses in the best SNN for each epoch of the best EONS run. Note: Given our encoding/decoding scheme, the minimum number of neurons is 6.
• Figure 5: Violin plot of circliness in RSSim across 100 seeds for spawn locations (see \ref{['eq:starting_region']}). The vertical bar denotes the mean circliness and the dots denote the best overall circliness achieved during training.