A Self-Organized Algorithm For Aggregation In Swarm Robotics

In nature, swarm animals take advantage of aggregation behavior for preventing from predator, keeping central temperature, making decisions on selecting shelter and so on. In swarm intelligence and swarm robotics that are inspired from the swarm animals, the aggregation of swarm robots is considered as one of the most basic collective behaviors as well as the fundamental prerequisite in many swarm robotic applications. However, as individual robot has limited ability of sensing and communication in swarm robotics, many locally optimal aggregates always occur during the self-organized aggregation process. As a result, it is difficult to generate a globally optimal aggregate.To transfer the locally optimal aggregates into one globally optimal aggregate, that is, to make all the robots aggregate together, this paper presents an aggregation algorithm based on linear timer. Inspired from the amplification mechanism of signals leading the aggregation of swarm robots, the key idea of my algorithm is: each individual robot in an aggregate has a lifetime; the lifetime is linear proportional to the aggregate size, where the proportionality factor is called linear factor K; after the lifetime is over, the robot will leave from the aggregate in order to search for other aggregates; therefore, individual robot prefers to leave from small aggregate and to stay in large aggregate; then as time goes on, all the robots are expected to finally aggregate together.This paper evaluates the linear timer based aggregation algorithm through simulator and real robots. The experiments include three parts as follows.(1) Experiments based on simulator are done to evaluate the linear timer based aggregation algorithm. The simulation results reveal that the linear timer based aggregation algorithm is convergent and scalable. Furthermore, the performance of the liner timer based aggregation algorithm is affected by the linear factor K. When the robot density is low, the performance will be improved by increasing the value of K; when the robot density is high, the performance will be improved by decreasing the value of K.(2) The linear timer based aggregation algorithm is successfully implemented on real robots. This experiment shows that the linear timer based aggregation algorithm can be used in real robots. Moreover, as the sensing ability of real robot is more limited than that of the robot model in simulation, the successful implementation of the algorithm on real robots shows that the key mechanism of the algorithm is more pervasive and is not limited on specific robot model.(3) Comparison experiments between the linear timer based aggregation algorithm and the Probabilistic Finite State Automata (PFSA) based aggregation algorithm are done. The selected two PFSA based aggregation algorithms are the algorithm without neighbor feedback and the algorithm with neighbor feedback. The experiment results show that when the total number of robots is large, the linear timer based aggregation algorithm can produce larger aggregate than the two compared PFSA based algorithms during the same time.