Real-Time Implementation of Randomized Model Predictive Control for Autonomous Driving

Model predictive control (MPC) using randomized optimization is expected to solve different problems. However, it still faces various challenges for real-world applications. This paper attempts those and demonstrates a successful implementation of MPC on the autonomous driving radio-controlled (RC) car. First all, sample generation technique in frequency domain discussed. prevents undesirable randomness which affect smoothness steering operation. Second, proposed implemented Graphics Processing Unit (GPU). The GPU acceleration calculation speed at problem sizes also presented. results show improved performance computational that was not achievable CPU based implementation. Besides, selection parameters usefulness scheme demonstrated by both simulation experiments. In experiments, 1/10 model RC car used collision avoidance task driving.