Neuroinspired Algorithms for Swarming Applications

Unmet Need
Swarm Intelligence is a relatively new branch of Artificial Intelligence that models the collective behavior of social swarms in groups of animals (ants, bees, birds). Swarming applications are agent-based algorithms that are capable of developing robust, self-organized solutions to emerging critical needs in modern day society. A major challenge in swarm control is maximizing communication efficiency for large swarms of small-scale robotic agents, allowing them to cohesively navigate real-world environments and perform tasks. With the rise in research and development of swarming applications, new approaches to autonomous but human-guided control of many agents are in high demand.

Technology Overview
This technology is a novel methodological framework for local, bottom-up control of spatially distributed mobile autonomous systems. The research team discovered a new class of spatial neurons (coined “phaser cells”) and applied the neural codes of those cells to an oscillatory learning model of distributed swarms. The resulting algorithm accounts for swarm movement as a neural interaction of the timing (oscillatory phase) and learning (phase-coupled association) between agents, which is analogous to neural circuits of spatial cognition in mammals. The resulting swarm behavior is adaptive and resilient, allowing swarms to dynamically replan around obstacles or other changes in the environment, unlike conventional top-down methods.

Stage of Development
This technology has been successfully prototyped for several use-cases.

Publications
JD Monaco, GM Hwang, KM Schultz, K Zhang. “Cognitive swarming in complex environments with attractor dynamics and oscillatory computing.”  arXiv:1909.06711 (15 Sep 2019).

Category(s):