Study On Obstacle Avoidance And Dynamic Cooperation Control Methods Of Behavior-based Robotics

Regarding the autonomous robot system research, the early research mostly was conducted under known and structural circumstances, and the robot had the precise understanding in detail before it makes movement. However, with the increasing demand for the robot intelligence, researching on non-structure environment becomes the mainstream. More and more researches no longer focus on the quantity of the environment knowledge obtained. Besides, the traditional control method usually requires the precise digital model in between input and output for controlled objects, which is very difficult normally. Although it can conduct the analysis description regarding the dynamic of robot itself, it is hardly to obtain the interaction between non-structure environment and robot through the sensor and executor. Under the non-structure environment, the lack of the accurate and complete knowledge of the environment also limits the application of independent robot on the standard control system. Under the dynamic and complex environment, the robot system needs to overcome the dynamic and uncertainty of environment in order to accomplish complicated mission.therefore, it is urgent to have the intelligence control system which is able to learn and coordinate from the experience under the uncertain environment.Regarding the above issues, this thesis combines national 863 high tech sponsored project, from the perspective of self-independent intelligence system, the thesis develops the deep research at aspect of the interaction between behavior control of lunar rover and intelligence system under non-structure complex environment.it conducts the preprimary theory research toward the our national lunar survey the second project. The specific research includes the following aspects: In this thesis, a control method of a mobile robot in unknown environments is discussed. The thesis proposes a new hierarchical fuzzy behavior-based control architecture to solve the navigation problem. An active autonomous ability in decision by itself is very important in unknown circumstances such as lunar environment. While in early research on behavior-based robotics, the robot has a little a bility on this aspect. On one hand, a prarell behavior-based architecture is adopted in our method by using the information extracted from its sensors to give the robot more ability on activity and reapid responses. On the other hand, the plan level can give the robot more macroscopical control such as giving sub goals, supervising motion of the robot and giving optimal paths. The architecture proposed in this thesis has an active robust ability in local behaviors and long-range optimal planning.The core theoretical foundations on designing of the Compound Zeno Behavior in hierarchical hybrid behaviors are presented in this thesis. Zeno behavior in behavior-based system, which is two or more deterministic behaviors make some number of transitions in finite time, can cause the motion speed and rotational velocity of the mobile robot change precipitately. Furthermore, the sharp shift of different behaviors will even exacerbate the absolute position errors. This thesis presents a new obstacle avoidance methodology to eliminate the negative effect of the sharp shift of different behaviors. Theorems are improved to prove the feasibility of the algorithm. Furthermore a smoothness estimation method is discussed to give an appraisement of the robot navigation trajectory. After that, the thesis focuses on producing emergency behaviors that lead the mobile robot navigation around the obstacles without sharp shift of different behaviors. Simulation results illustrate the good performance of the control algorithms. BP Neural networks are used to select and fusion different behaviors. And the methods presented in this thesis can control the robot reaching the goal without collision with the obstacles, making the smoothness of the trajectory, and showing the good performance of the architecture. A cycle diagnosis navigation strategy is proposed in this thesis according to the inspection and managment of gorge effect during the navigation of the robot. By the guidance of the cycle diagnosis navigation strategy, the robot can navigate out of from the concave terrain, which improves the robust of the autonomous systems when dealing with the complex environment.For the conscious cooperation issues of multi-robot while executing tasks, two types of task allocation algorithms are presented on the basis of classifying task allocation problems: GA based task allocation method and dynamic decision thehod. Genetic algorithm is proposed in this thesis to solve middle large scale of static task allocation problems. The solving time is rapid enough and has applicability on finding the optimal solution when solving the middle scale of static task allocation problems. To solving the uncertainty dynamic team combination problems, a cooperative task decision mechanism based on behavior profit evaluating value is proposed in this thesis. The simulation results indicate that the method can satisfy the real time optimal autonomous robust request of the system. A good performance of eliminating the deadlock of the robots is validated by different simulation experiments.Furthermore, a 3D lunar rover simulation environment is developed based on VRML techniques to give the validity of the proposed methods in this thesis. Control architecture and sub systems are designed in detail to simulate the navigation situations in unstructed environment. Plenty of experimentations are presented to give some useful results. The simulation environment is such a good flat in studying the autonomous control methods of behavior based autonomous mobile robot.At the end of this thesis, summarizations are given to look forward the future works.