An intelligent-optimal predictive controller for path tracking of Vision-based Automated Guided Vehicle

An original path tracking controller for a vision-based automated guided vehicle (V-AGV) is presented in this paper, containing a hybrid algorithm of intelligent-optimal predictive control and an actual embedded system. For a kinematics model in local coordinate, the aim of path tracking is to eliminate two path errors of distance and angle. The optimal predictive control via LQR is applied when path errors are small, which can reduce theoretically two path errors to zero synchronously and smoothly in the minimum steps. The intelligent predictive control imitating human driving is used when path errors are large, which can decrease two path errors quickly, and convert large errors into small errors harmoniously. The most remarkable features of the hybrid algorithm are realizable velocity output and excellent control performance in different conditions of path errors and velocities. An advanced embedded controller, consisting of two microcontrollers of DSP TMS320DM642 and ARM LPC2200, is developed to execute the hybrid algorithm of path tracking. The results of numerical simulations and tests on a practical V-AGV system both demonstrate the effectivity and advantages of the algorithm and controller presented in this paper.