A New Approach for Mapping of Soccer Robot Agents Position to Real Filed Based on Multi-Core Fuzzy Clustering
Seyed Omid Azarkasb, Seyed Hossein Khasteh · 2021
Mapping the position of soccer robot agents to a real field, is one of the essential issues in the practical implementation of scientific contributions in this context. The lack of a proper assignment affects the scientific implementation of many subjects, such as routing, obstacle avoidance, and robot guidance. For this reason, the use of a clustering method is proposed in this article. Upon the entrance of a new agent, its position is mapped to the real field based on the clustering algorithm. After this mapping, the system begins to work according to the position of the agents, which is defined as the position of the centers of the clusters, as well as the rules defined in the knowledge-base. Considering the unknown and dynamic environment of the robot, some objects inherit common traits from multiple clusters. One reasonable solution for considering the cluster overlaps is to assign a set of membership degrees to each of them. Multiple membership degree assignments result from the fuzzy nature of the clusters. Due to the reduction of segmentations and the shrinkage of the search space, fuzzy clustering generally faces less computational overhead, while the identification and handling of vague, noisy, and outlier data also become much easier in them. The approach of the proposed method is based on the feasibility ideas and uses multi-core learning to identify clusters with complex data structures. The feasibility score of each data represents the percentages of the properties that data inherits from the clusters. Automatically adjusting the weights of the cores in an optimization framework, the proposed method avoids the damage caused by problems such as adopting inefficient cores, or irrelevant features.