Translated Abstract
The intelligent mobile robot is a multi-functional integrated system which can perceive the environment, make dynamic decision, plan and control its behavior. Comparing with the fixed robot and manipulator, the range of application and function of the mobile robot is greatly expanded and improved, therefore the mobile robot is used widely in such aspects as industry, national defence, service industries, and receives more and more attention too. Among all types of the walking mobile robots the wheeled mobile robot can be “put in an important position” because of its simple mechanism, energy-conservation, less degrees of freedom, easy control, higher practicality.With the constant development of computer, communication and control technology, people are looking for various kinds of methods to spare no effort to utilize the moving carriers such as mobile robot or autonomous vehicle, etc. to complete some repeating, simple, dangerous tasks. If mobile robot wants to realize intelligent and autonomous motion, it must possess self-position technology, intelligent motion control technology and self-learning mechanism. Combining with existing experimental conditions some research about mobility, flexibility and intellectualization of the mobile robot is systematically studied in this dissertation as follows:1. The open intelligent control system of the mobile robot for experiment is build up in which the hardware platform adopts PC+DSP mode with easily realizing, strongly functioning and reliably controlling. Correspondent to the hardware system, the software system is developed. The key subsystem of the mobile robot — the position system is studied and its error source is analysed, then the corresponding correcting method is proposed, the experimental result shows the validity of the method. For reducing the error of the relative localization measurement the position information provided by dead reckoning model using multi-sensor data fusion with Kalman filter algorithm which improve the precision and reliability of the position system, finally the laser navigation system is used to verify its effectiveness.2. The structure characteristics and models of the wheeled mobile robot are summarized. A common method using coordinate transformation is put forward to solve kinematics modeling of the mobile robot with conventional wheels including drivable and steerable wheels which set up the direct kinematics model of the mobile robot under desire motion constraints. In addition, the inverse kinematics model is deduced. Taking 3-DOF mobile robot with 2 drivable and steerable wheels as example, the kinetics model is derived with Kane method which provides the theoretical foundation for designing the tracking controller.3 The 3-DOF mobile robot with 2 drivable and steerable wheels is a strong coupling general non-linear system whose motion following control is solved by the inverse systems approach. On the premise of proving its invertibility the pseudo-linear system which realizes the decoupling between input and output of the system is implemented. According to the neural network th-order inverse systems method the controlled non-linear mobile robot system is approximately decoupled into a number of independent SISO linear subsystem by using BP neural network and dynamic integrator. Then developed linear system theory is used to design an additional feedback linear controller to successfully realize the independent position and orientation tracking of the mobile robot. The simulation result shows its good decoupling and tracking performance.4. In order to overcome the influence factors caused by the structure biases of mobile robot, measuring noise, environmental disturbance, etc. in practical application, a tracking controller based on fuzzy logic method is presented. The control rules and parameters of the fuzzy controller are optimized with an improved genetic algorithm which realizes the optimization of T-S fuzzy controller of the general complicated system. The tracking results of the simulation and test demonstrate the better performance of the GA-fuzzy controller than that of the traditional PID controller.5. An adaptive fuzzy neural network controller is presented to improve adapting and self-learning ability of the autonomous mobile robot. In real-time control process, the weight coefficients of the critic network and parameters of the actor are adjusted based on inner reinforcement signal. The results of the simulation and test indicate that the adaptation ability and tracking performance are improved.6. Aimed at problems of autonomous navigation for mobile robot with vision servo system, a double population genetic algorithm with the surface-stripe model matching object-feature based applied to recognize navigate base-line is proposed. On the basis of the inverse perspective mapping theory the pose bias which can be corrected by fuzzy-reasoning based self-tuning PID controller is analysed. Because the input raw image—grey scale image is treated directly, the time of image processing and explanation is shortened greatly. The method guarantees the real-time control requirements of the mobile robot. Finally the real image recognition and vision navigation experiment verifies the validity and feasibility to recognize the object with shape feature and robustness against environmental noise.
Corresponding authors email
