Evolving Behaviour Trees for Swarm Robotics

Abstract

Controllers for swarms of robots are hard to design as swarm behaviour emerges from their interaction, and so controllers are often evolved. However, these evolved controllers are often difficult to understand, limiting our ability to predict swarm behaviour. We suggest behaviour trees are a good control architecture for swarm robotics, as they are comprehensible and promote modular reuse. We design a foraging task for kilobots and evolve a behaviour tree capable of performing that task, both in simulation and reality, and show the controller is compact and understandable.

Behavior Trees as a Control Architecture in the Automatic Modular Design of Robot Swarms

“…The results show that behavior trees can be effectively used to control the robots of a swarm and that the control software generated is human-readable. Our approach differs both conceptually and methodically from method proposed by Jones et al [24].…”

“…Jones et al [24] proposed an automatic design method for robot swarms in which the control architecture of robots is a behavior tree. To the best of our knowledge, that is the first and only application of behavior trees in swarm robotics.…”

Optimization of Swarm Behavior Assisted by an Automatic Local Proof for a Pattern Formation Task

“…Evolutionary algorithms can search through vast solution spaces and discover solutions to complex problems, and are thus a popular approach to dealing with the intricacies of swarm robotics and extracting valid local behaviors [6,14]. They have been used for numerous architectures, including: neural networks [3,11], state machines [7], behavior trees [12], and grammar rules [5]. When applied to swarms, the following issues typically arise:…”

Robust distributed decision-making in robot swarms: Exploiting a third truth state

“…We employ a Kilobot simulation environment [11] which captures many of the physical properties of a Kilobot swarm including motion, collisions, and communication between robots. The simulator also uses the same API 2 as used on the actual robots which makes it easier to transfer code from simulations to the real world.…”