A Self-Organized Algorithm For Size Estimation In Swarm Robot System

With robot technology development and practical need, as a type of multi-robot system promoting self-organize, scalability and robustness, swarm robot has become an attractive research area home and aboard. The number of robots is a fundamental parameter in swarm robot system, and estimation of robot number is required by some swarm robot control algorithms. The traditional distributed size estimation algorithms expo low-robust and high communication requirement when migrated to swarm robot system. As a result, a kind of size estimation algorithm for swarm robot has high research value.This paper proposed a pheromone-based size estimation algorithm. The basic idea of algorithm is: constant quantity of virtual pheromone spread in the swarm robot system; after each node gets same quantity of pheromone, robot can estimate the system size based on the density of pheromone. In our algorithm, pheromone spread through a gossip-based method with session mechanism, which guarantees reliability and scalability of communication. In the spread process, robot records the history exchange amount with neighbors. So if pheromone is lost in robot crash or communication issue, robots can restore local pheromone to keep the total amount unchanged. In the role assignment, a self-organized competition mechanism guarantees there's only one source node of pheromone in the group, which makes the algorithm available in the dynamic scenario that group keeps being reorganized.My main study work includes:1. Design a group size estimation algorithm for swarm robot: Design a group size estimation algorithm suitable for swarm robot, and solved three main problems: reliable spread mechanism for pheromone in full distributed environment, automatically retrieve mechanism when total quantity of pheromone changed and self-organized role assignment in swarm robots.2. Implement and analytic algorithm in simulation environment: Implement our algorithm and comparison algorithm in a simulator; design the evaluation scenarios; gather and compare experiment data; evaluate accuracy, dynamic performance, robust and scalability of algorithm.3. Implement algorithm on real robot: Migrated the algorithm to real robot to verify the correctness in real environment.From the result of the experiment on simulator and real robot, our algorithm can estimate group size accurately. Compared with other algorithm under same condition, our algorithm shows better dynamic performance, robust and scalability. Our algorithm is full-distributed, and can be applied to swarm robot system consisted quantity of robots.