Abstract:
Swarm Robotics is the study of swarms of small simple locally behaving mobile robots and their emergent behaviors as a swarm of intelligent species. The inspiration of these studies comes from nature swarms of ants, fish and birds colonies. Many areas of swarms including sensing, localization, mapping, path planning, interference and behavior modeling are being studied currently.
Path planning being one of the old areas of researches has already been studied in great details for static and moving obstacles. The introduction to intelligent obstacles that react to the robots motion are new addition to the research areas. Swarms of robots are numbers of robot trying to accomplish collective or individual tasks. Hence the area of path planning has taken a new dimension. “Reciprocal Velocity Obstacles” is one of the approaches addressing the same problem. It though lacks a mechanism for decision symmetry breaking and biasing. Biases RVO is one attempt to do so, yet the attempt is more of an inspirational concept as it is not much generic in nature and must be modeled for every different scenario.
Getting inspiration from swarm congestion control techniques and ant foraging models, two approaches we tried for implementation for asymmetric biasing and fast solutions to swarm problems. The first one is a much global behavior technique in which the robots try to avoid the center of mass of the whole swarm weighted by inverse of the radius of gyration of the point masses of the robots. It worked well for simulations possible congestion at one point but lacked better results in situations of multiple congestion zones. The second one is a much local approach. In this approach the robots try to follow their most friendly neighbors, weighted by the distance from their respective goal positions. In this way, it was observed that the robots followed a much humble path when they were away from their goals and became ruder as the goals came nearer. The technique proved results in both local and global congestion zones. The efficiency benchmarking for different scenarios varied from 30% in mobile robot pick and place situations to 50% in a symmetric circle antipodal goals problem.
