Rough Set Theory Based Behavior Decision-making Of Patrol-robot

Environment perception, behavior decision-making and action are the three important parts of robotics architecture. The behavior decision is the basis of harmony and control of behaviors and the key aspect of robot's intelligence. In this thesis, the behavior decision-making of robot is studied based on the Patrol-robot, which is a kind of autonomous mobile robots to perform patrolling and security-monitoring.Firstly, the researching state of autonomous mobile robots and Patrol-robots in our country and abroad is analyzed. The framework of decision-making based on rough set and multi-decision-making table is proposed, for the advantage of rough sets in learning and decision-making.Secondly, through analyzing the operating environment in which Patrol-robot runs, the three-level environment model: global model, local model, instantaneous model is developed based on topologic map. Road detection plays a significant role in the environment perception system of Patrol-robot. Unstructured road detection algorithm based on rough set is proposed according to the advantages of rough set in feature-learning and the test validates its robustness and reliability..Thirdly, the behavior based task control architecture is proposed and the behavior decision-making model based on multi-decision-making-table is developed. The multi-decision-making-table based implementation of behaviors and navigating behavior decision-making are studied. In the implementation of behaviors, the behavior of objective monitoring is taken as an example to expatiate on the procedure of acquiring knowledge of decision making and the method of implementation of behaviors using knowledge database. In the navigating behavior decision-making system, by analyzing the states, relationship and condition of conversion of Patrol-robot during patrolling, the finite-state-machine of Patrol-robot is formed. The rule set of decision-making information system is formed by abstracting the conversion condition of finite-state-machine. The simplest set of decision-making rules is obtained by attribute reduction and rules abstracting algorithm. The knowledge database of behavior decision-making is established and applied in the effectiveness experiment of patrolling navigation.At last, the experiment of navigating behavior decision-making based on the knowledge database is actualized and the analysis based on result data of the experiment is given. In the experiment of navigation, the robot is reactive quickly during navigating by using the navigation knowledge database and got to the goal at last. The decision system is proved available in the whole system of patrol-robot.