The Information-Cost-Reward framework for understanding robot swarm foraging

Abstract

Demand for autonomous swarms, where robots can cooperate with each other without human intervention, is set to grow rapidly in the near future. Currently, one of the main challenges in swarm robotics is understanding how the behaviour of individual robots leads to an observed emergent collective performance. In this paper, a novel approach to understanding robot swarms that perform foraging is proposed in the form of the InformationCost-Reward (ICR) framework. The framework relates the way in which robots obtain and share information (about where work needs to be done) to the swarm's ability to exploit that information in order to obtain reward efficiently in the context of a particular task and environment. The ICR framework can be applied to analyse underlying mechanisms that lead to observed swarm performance, as well as to inform hypotheses about the suitability of a particular robot control strategy for new swarm missions. Additionally, the informationcentred understanding that the framework offers paves a way towards a new swarm design methodology where general principles of collective robot behaviour guide algorithm design.

The Importance of Information Flow Regulation in Preferentially Foraging Robot Swarms

“…A continuous-space experimental arena, identical to that in [22] and containing a central circular base surrounded by circular worksites, was created in the ARGoS simulator [18]. An experimental environment was characterised by the number of worksites, N W , and worksite distance from the base, D ∈ {5, 9, 13, 17} m. There were two types of environment (Figure 1a,b): -HeapN W : N W ∈ {1, 2, 4} high-volume worksites evenly distributed around the base at a distance D from the base edge.…”

“…Two types of foraging tasks were investigated, as in [22]: -Consumption: Worksites represented "jobs" that could be completed at the worksite locations. A robot that was at a worksite gradually depleted its volume, increasing the swarm's total reward at the rate of 1/400 units per second.…”

“…Worksite volumes were replenished after each change. The value of T C , as well as the total simulation time, T , were set as in [22], so that the environment changed 10 times in the slow variant and 20 times in the fast variant and so that a swarm could deplete around 50% of total worksite volume in each slow change interval. For example, T C = 45 min and T = 7.5 h for 50-robot swarms.…”

“…The robots could communicate with each other using a range and bearing module with a signal range of 5 m. We have previously described the robot model in [20]. There were two types of robot swarm, that we parametrised for the best performance in a series of environments [22]: -Broadcaster (Figure 2a): Robots left the base immediately at the beginning of an experiment to start scouting for worksites. Upon discovering a worksite, a scout started working on it, meaning that it either started depleting its volume (in the Consumption task) or started travelling between the worksite and the base to gradually deposit resource (in the Collection task).…”

“…In Bee swarms, robots exchange information in a designated area, that they return to periodically. Our previous work [20,22] suggests that Broadcaster swarms outperform Bee swarms in many foraging environments due to their ability to respond to environmental changes faster, but that Bee swarms are more suitable during central-place foraging in environments with a low worksite density, since their recruitment strategy allows the robots to share information with each other relatively quickly. Here we show that the rapid spread of information through Bee swarms damages their performance when individuals preferentially choose worksites with higher utility, since most of the swarm may tend to concentrate on a single worksite, increasing the negative effects of congestion.…”