There are growing applications of multiagent systems in which each mobile agent moves towards its destination while avoiding others, e.g. robots in warehouses, self-driving cars, and flying drones. In these systems, it is desirable that each agent moves quickly, smoothly, and safely. However, simple methods for satisfying these three functions simultaneously have not been developed. To address this challenge, we capture the essential control mechanism for achieving multiobjective tasks, drawing inspiration from pedestrian flow. We propose a decentralised control scheme, an extension of the social force model, which is a simple model of pedestrian flow, wherein agents can avoid other agents based on the prediction of their future motions. Through simulations, we demonstrate that the proposed control scheme enables agents to move quickly, smoothly, and safely.
|Physica A: Statistical Mechanics and its Applications
|Published - 2021 Jun 15
- Decentralised control
- Pedestrian flow
- Social force model