Research On Automatic Design Method Of Behavior Logic For Swarm Robotics

Nowadays,the technology of unmanned cluster has become an important focus to master the war situation.The application of intelligent multi-robot system has become the key technology to generate important combat effectiveness and grasp the battle rhythm.Among them,swarm robotics system is the key problem that the development of unmanned cluster technology needs to break through first,but also the foundation of multi-unmanned cluster cooperative warfare.Due to the limited ability of single robot in a swarm and the requirement of ability in complex battlefield for swarm robot,traditional methods are difficult to provide effective support for automatic design of behavior logic for swarm robotics.Therefore,based on the highly explainable behavior tree representation method and the explainable evolutionary algorithm,this thesis aims at the automatic design method of behavior logic for swarm robotics in the face of complex and changeable multi-task scenarios and other challenges,and carries out the following four key technologies:(1)Facing the demands of autonomous and intelligent swarm robotics,in response to the challenges of the lack of universal and flexible automatic design methods applicable to different swarm robotics,based on the behavior nature and structural characteristics of swarm robotics,this thesis construct the automatic design framework of behavior logic for swarm robotics.This framework provides feasible theoretical framework support for the optimization design of the behavior logic of swarm robotics.(2)Facing the demands of interpretability and easily-optimized of the representation method of behavior logic for swarm robotics,in response to the challenges of flexibly completing a variety of complex tasks within the scope of the ability of swarm robotics,and the optimization problem of behavior logic,this thesis proposes a fixed structure behavior tree representation model of behavior logic for swarm robotics,which realizes the design of general and interpretable representation method of behavior logic.In addition,a matrix model of the fixed structure behavior tree is constructed,which lays a foundation for the optimization algorithm design of behavior logic.(3)Facing the demands of fully distributed structure characteristics of swarm robots with limited ability,based on the behavior logic of swarm robots represented by matrixized behavior tree,in response to the challenge of limited individual perception and computing capability of swarm robotics,this thesis designs a new evolutionary algorithm for the behavior logic of swarm robotics.In this algorithm,fully considering the representation of behavior logic,the efficiency of evolution and the requirements of evaluation of behavior logic,various genetic manipulations,a conditional filtering mechanism and a behavior logic evaluation method through simulation experiments are designed.Therefore,an appropriate behavior logic for swarm robotics can be obtained through this algorithm.(4)Facing the demands of experimental verification of scalability,interpretability,adaptability and superiority of the proposed method,multiple groups of experiments are carried out for a swarm in a task scenario based on MATLAB simulation software,Kilombo simulator,and Kilobots platform.The performance of the same swarm robotics in task scenarios with different complexity of tasks,population sizes and algorithms is discussed,which lays a solid foundation for the further application of the proposed automatic design method of behavior logic for swarm robotics in real systems.