Dge and also the parameter tuning time. The sensible weighting matrices and
Dge as well as the parameter tuning time. The sensible weighting matrices and had been additional revised pre-trained datum worth of the weighting matrix, it may matrices applied in non-RLMPC for RLMPC, as indicated in Equation (58). The weighting significantly reduce the parameter tuning time. The the operator were matrices as and Rn were further revised for Equathat had been tuned bypractical weighting the same Qn the simulation case indicated in RLMPC, as indicated in Equation (58). The weighting matrices applied in non-RLMPC that were tion (53). tuned by the operator were thethe path PF-06873600 manufacturer tracking resultscase indicated in Equation (53). For scenario 1 experiments, exact same because the simulation of MPC and RLMPC are shown For scenario 1 tracking errors path tracking final results are indicated in Figure 11. The in Figure ten, and theexperiments, theof MPC and RLMPC of MPC and RLMPC are shown in Figure 10, and theresults were quiteMPC and RLMPC are indicated in Figure 11. results line path tracking tracking errors of equivalent for the aforementioned simulation The line path in Figures five and six. The human-tuned MPC represented simulation benefits shown shown tracking final results had been very similar towards the aforementioned some oscillation when thein Figures 5 the six. The human-tuned MPC represented some oscillation error immediately after the 70th EV reachedand line path. Nonetheless, the RLMPC exhibited a smallerwhen the EV reached the line sample. path. Nonetheless, the RLMPC exhibited a smaller error right after the 70th sample.Figure 10. Trajectory comparison MPC and RLMPC in situation 1. Figure ten. Trajectory comparison ofof MPC and RLMPC in situation 1.For the scenario two experiments, the path tracking final results of MPC and RLMPC are shown in Figure 12, plus the tracking errors of MPC and RLMPC are indicated in Figure 13. It was apparent that the RLMPC outperformed the tracking error compared to the humantuned MPC. To supply a confident and quantitative error evaluation, each of the experiments had been performed three instances for the overall performance comparison, as indicated in Table 4. Table four shows the relative statistical data of averaging the values with the 3 trials. Each of the typical RMSEs have been significantly less than 0.3 m, plus the maximum errors had been significantly less than 0.7 m.Electronics 2021, 10,18 ofThe overall results showed that the RLMPC and human-tuned MPC Thromboxane B2 manufacturer followed exactly the same ronics 2021, ten, x FOR PEER Critique trajectory nicely. However, with well-converged parameters, RLMPC had much better efficiency than MPC tuned by humans when it comes to maximum error, average error, regular deviation, and RMSE.Figure 11. Tracking error comparison of MPC and RLMPC in Situation 1.Figure Tracking error comparison of MPC and Scenario in Figure 11.11. Tracking error comparison of MPC and RLMPC inRLMPC1. Scenario 1.For the situation 2 experiments, the path tracking benefits of MPC and shown in Figure 12, and the tracking errors of MPC and RLMPC are indica 13. It was apparent that the RLMPC outperformed the tracking error com human-tuned MPC. To provide a confident and quantitative error evalu experiments were performed three occasions for the functionality comparison, a Table four. Table 4 shows the relative statistical information of averaging the worth trials. Both of the average RMSEs have been much less than 0.three m, along with the maximum er than 0.7 m. The overall benefits showed that the RLMPC and human-tuned M the same trajectory nicely. Having said that, with well-converged parameters, RLM performance than MPC tuned by humans in terms of maximum error, a regular deviation, and RMSE.For t.