| dc.description.abstract | This dissertation proposes the methods of localization and body attitude balance, which were adopted in two hardware devices, mobile robot and hexapod robot.
In terms of the indoor service mobile robot design, in addition to target object image recognition, target object gripping, and obstacle avoidance functions, the research of mobile robot focuses on the relative position location system of the robot. The localization method uses the values of two sensor modules, which are gyroscope and magnetometer, to correct the current rotation direction angle of the robot. The angle correction is divided into three parts: the first part calculates the angle values of the gyroscope and magnetometer that are mounted on the robot; the second part obtains the error characteristics between the sensor modules and the actual rotation direction angle of the robot; the third part uses the data of these error characteristics to design a fuzzy rule base and Kalman filter parameter, and uses them to eliminate errors in order to obtain more accurate direction angle. These error characteristics can be described as regular and irregular errors, where the former can be eliminated using the fuzzy theory, and the latter can be eliminated using the Kalman filter theory. The contribution of this dissertation is the error correction method for sensors and actual rotation angle of the robot, where the specified path, actual path, and calculated path can approximate to each other, thus, implementing accurate localization of the intelligent robot. The experimental results show that the combination of fuzzy compensation and Kalman filter is an accurate correction method.
In addition, in order to validate the feasibility of the application of the improved Central Pattern Generator (CPG), as proposed in this research for robot body attitude balance control, the hexapod robot was used as the hardware of theoretical examination. The CPG controller is a type of distributed control method; it imitates the regular motion control mechanism of the organisms in the robot gait motion design, the mutual transmission and effect of low-level neural cells generate regular and periodic signals, and the signal is corrected by external information or cerebral irritation, and synthesized into the final motion mode. In terms of the gait design, each leg of the robot is controlled by an improved CPG controller, connected with the CPG of the other robot legs, and different motion gaits are generated by different connected modes. The Matsuoka neural oscillator as the basic composition unit of CPG, and proposes a new CPG architecture: in the annular three-link double neuron CPG architecture in charge of oscillator phasing, an external neural oscillator is added, which is in charge of adjusting the amplitude of oscillator in order to control the treading depth of the robot legs. The overall control architecture uses an accelerometer and a gyroscope to obtain the real-time robot body attitude, while the tilt angles of the leg directions are separated as feedback signals imported into the CPG to change the amplitude. It is compared with the reference oscillator of fixed amplitude to generate the leg height reference signal, which can balance the body. Afterwards, the control signal is converted by the trajectory generator into the track of the robot leg action. The actual servo motor rotation angle is obtained from this trajectory by inverse kinematics to control the motor rotation angle. Thus, the robot can move forward, and instantly restore horizontal body attitude when walking on rugged terrain. The experimental results show that the gait design method proposed in this research enables the hexapod robot to walk smoothly on rugged terrains. | en_US |