Design of a Fuzzy Logic Controller for Skid Steer Mobile Robot

The control problem of four-wheeled skid steering mobile robots is quite challenging mainly because the skid steering system is an underactuated system and its mathematical model is highly uncertain. Skid steering configurations employ a differential-drive technique in which the wheels rotation is limited to around one axis and the lack of a steering wheel causes the navigation to be determined by the change of speed in either side of the robot for turning. Equal speed in both sides causes a straight-line motion. However, the implementation of the dead reckoning technique on skid-steer mobile robots will limit the precision of current robot’s position because skid-steer configuration intentionally relies on wheel slippage for normal operation and this possesses some difficulties when implementing motion control using the odometric system. The thesis describes the design of a fuzzy logic controller to compensate the dead reckoning limitation and implementation on a skid-steer mobile robot. The fuzzy controller has two inputs (angle error and distance), two outputs (translational and rotational speed) and 14 rules. These inputs are computed from the dead-reckoning method that is totally reliant on the odometry readings and data are fuzzified to be the inputs of the fuzzy controller. The outputs are the analogue voltages to the left and right motors, which drive the mobile robot. For simplicity, membership functions consisting of triangular and trapezoid shapes have been adopted. The membership functions of the fuzzy sets are chosen by trial-and-error based on experimentation. The heuristic rules control the orientation of the robot according to the information about the distances from the desired positions. The crisp output values from the fuzzy logic controller are decoded and fed into a decision module where the ratios of both sides motor voltage are determined for every smooth change in speed of the motors. To facilitate the implementation of control system, real-time execution is done in an indoor environment. Data acquisition is done in a LABVIEW and a MATLAB control algorithm is called in LABVIEW. A real mobile robot, PUTRABOT2 was used to conduct the experiment. Performance evaluation is observed from the accumulated error in orientation and its trajectory obtained after mapping the information gathered from the real world via odometry sensors. Few features such as the rise time, settling time and peak time of the output responses are analyzed. Comparisons are made between fuzzy logic and PD controllers. Comparative results among these two controllers indicate the superiority of the fuzzy approach with the ability to minimize the position and orientation errors. Moreover, the trajectory accuracy is very high and more reliable in the presence of unreliable odometry readings.

Preview PDF ABS__FK_2009_74.pdf Download (287kB)

Actions (login required)

View Item