Collective transport by caging in swarm robotics

Resumo

Collective transport, in swarm robotics, is based on the transportation of objects by swarms of robots characterized by having significantly smaller dimensions compared to the object to be transported. The set of robots of the swarm generally have the same architecture. Individually, the robots are able to perform simple functions, such as sensing, locomotion, and basic communication. However, cooperatively, they can perform complex task. Collective transport has diverse applications that encompass the transportation of both large-scale objects and nanoscale objects. Consequently, numerous studies have been driven by this topic, highlighting three  transport strategies: pushing, grasping, and caging. In this work, the caging strategy is adopted. It can be defined as the complete enclosure of the object. Allowing it a certain degree of freedom but preventing it from escaping the formation of robots around it. The main advantage of this strategy is the coordinated and cohesive progression of the transport since the forces applied to the object by the swarm complement each other and prevent significant deviations in its trajectory towards the planned destination. This work proposes a method to approach collective transport by caging, which operates in three stages: the search for the object by the swarm, the recruitment of swarm robots, aiming to position the robots around the object and enclose it, and the transport stage. The implementation of the proposed strategy is carried out in the CoppeliaSim platform, wherein the swarm is composed of Khepera III-type robots. This robot is chosen due to the arrangement of its sensors along  its perimeter, which allows for a wide view of the environment. The arena, which is the scenario where the simulations are conducted is discretized into different configurations. This space discretization has a significant impact on the performance of the search and transport stages. The performance evaluation of the search and recruitment stages are based on the execution time, while for the transport stage, in addition to the execution time, the normalized error of the object’s trajectory is evaluated for different arena discretization configurations and path lengths. The proposed method proves to be effective, as the caging is maintained throughout the entire path, thus ensuring the uniform transport of the object.