基于MCTS-HM的重型汽车多参数运行工况高效构建方法

Multi-Parameter Driving Cycle Construction for Heavy-Duty Vehicles Based on MCTS-HM

提出了一种基于蒙特卡洛树搜索(Monte Carlo Tree Search,MCTS)和启发式方法 (Heuristic Method,HM)的多参数运行工况高效构建方法。根据工况质量要求,选取表征行驶工况特征的速度、加速度和坡度参数相关指标,设计工况的目标函数;改进多个启发式方法,确保被作用后的工况序列仍可满足行驶工况动态转移特性;采用MCTS选择更适合的启发式方法,进而高效地收敛至多参数目标行驶工况。基于重型汽车实际采集的试验数据,多次分析结果表明,生成的多参数行驶工况所有特征指标相对偏差均可保持在设定阈值以内,表明生成工况与原始数据特征具备一致性;相比于纯随机启发式方法,该方法能快速构建多参数期望运行工况。

Construction of representative driving cycles is fundamental for evaluating performance of heavyduty vehicles(HDVs) and developing new vehicle technologies. Due to the unavoidable influence of frequent changes in road slope on performance evaluation and testing of HDVs, it is necessary to construct the multiparameter driving cycles which can be generated with high quality and efficiency. However, the random heuristic methods often violate the driving cycle characteristics, resulting in the low generation efficiency.Therefore this paper proposed an effective method to construct multi-parameter driving cycles based on Monte Carlo Tree Search(MCTS) and Heuristic Method(HM). According to the quality requirements for driving cycles, some parameters including velocity, acceleration and road slope were selected to form the cycle objective function. Multiple heuristic methods were improved so that the sequences in the driving cycles can still meet the dynamic transition characteristics. MCTS was used to select a suitable heuristic to efficiently converge to the target multi-parameter driving cycles. Based on the collected experimental data of HDVs,the analysis results show that the relative deviations of all the feature indices for the generated multi-parameter driving cycles are kept within the set threshold, which indicates that the characteristics of the generated driving cycles are consistent with those of the original data. Compared with the random heuristic methods, the proposed method can quickly construct multi-parameter driving cycles with higher quality.