Tracking Multiple Fast Targets With Swarms: Interplay Between Social Interaction and Agent Memory

Abstract

The task of searching for and tracking of multiple targets is a challenging one. However, most works in this area do not consider evasive targets that move faster than the agents comprising the multi-robot system. This is due to the assumption that the movement patterns of such targets, combined with their excessive speed, would make the task nearly impossible to accomplish. In this work, we show that this is not the case and we propose a decentralized search and tracking strategy in which the level of exploration and exploitation carried out by the swarm is adjustable. By tuning a swarm's exploration and exploitation dynamics, we demonstrate that there exists an optimal balance between the level of exploration and exploitation performed. This optimum maximizes its tracking performance and changes depending on the number of targets and the targets' movement profiles. We also show that the use of agent-based memory is critical in enabling the tracking of an evasive target. The obtained simulation results are validated through experimental tests with a decentralized swarm of six robots tracking a virtual fast-moving target.

Balancing Collective Exploration and Exploitation in Multi-Agent and Multi-Robot Systems: A Review

“…This allowed for an individual agent's contribution, and hence the system's overall exploration and exploitation dynamic to be quantified. Kwa et al (2021a) and Kwa et al (2021b) later identified that a metric based on an agent's position relative to a target was unable to function when used in a scenario where multiple targets are to be tracked, such as situations defined under the Cooperative Multi-Robot Observation of Multiple Moving Targets (CMOMMT) framework. This led to the introduction of an engagement ratio, thereby allowing quantification of a swarm's EED in the presence of multiple targets by finding the proportion of agents actively moving towards a target.…”

“…To track a fast moving target Kwa et al (2020a), Kwa et al (2020b), Kwa et al (2021a) and Kwa et al (2021b), initially used inter-agent repulsion to promote area exploration. When a target was detected, an agent and its neighbors would activate a PSObased tracking behavior, allowing the target's information to be exploited, facilitating target tracking.…”

“…By increasing the truncation probability, agents of a system could be made to use information from its memory more often, leading to a system-level EED that favors exploitation. Kwa et al (2021a) and Kwa et al (2021b) also made use of memory in their target tracking task. However, instead of changing the probability of an agent accessing information from its memory, the system's EED was tuned by changing the length of an agent's memory.…”

“…A similar repulsion scheme was adapted by Coquet et al (2019) for use in their dynamic target tracking swarm. Similarly, to prevent excessive agent clustering while carrying out a target tracking task Kwa et al (2020a), Kwa et al (2020b), Kwa et al (2021a), Kwa et al (2021b) employed an adaptive repulsion algorithm that gradually increased the strength of inter-agent repulsion while agents were exploring the area. This repulsion strength was also gradually decreased while agents were attempting to move towards a target.…”

“…Many MAS practitioners use the agents' movements and observed positions to determine the system's exploration-exploitation balance (see Section 2.2.1). For example, in a target search context, when agents are close together or traveling towards the same region of the area of operations, it can be assumed that they are exploiting a common source of information (Jordehi, 2014;Kwa et al, 2021b). Conversely, if agents are moving away from each other or in a random fashion, they can be assumed to be carrying out exploration.…”