Efficient Construction of Multi-Parameter Driving Cycles for Heavy-Duty Vehicles Through Monte Carlo Tree Search Heuristics

Assessing the effect of road grade on the performance evaluation and testing of heavy-duty vehicles (HDVs) requires the efficient construction of a high-quality multi-parameter driving cycle of HDVs. However, existing pure random heuristic methods fail to preserve the driving characteristics of the original driving cycles, resulting in poor-quality outputs. In addition, the randomness inherent in multiple heuristic approaches limits the search efficiency. To address these issues, this study proposes a novel Monte Carlo tree search heuristic method (MCTSHM) for efficiently constructing multi-parameter driving cycles of HDVs. First, a satisfactory criterion model was used to design the objective function for the multi-parameter driving cycle, ensuring the evaluation indices satisfy given constraints. Next, heuristics were designed to maintain the dynamic transition characteristics of driving cycles. An improved Monte Carlo tree search was conducted to efficiently select heuristics more suited to the current driving cycle. Finally, the expected driving cycle was returned by combining heuristics with the Monte Carlo tree search using a hyper-heuristic architecture. The experimental results were analyzed using driving data collected from HDVs. The relative deviation of all characteristic indices between the generated and original cycles remained within 10% of the set threshold, indicating high similarity to original database. In comparison with a purely random selection of heuristics, the MCTSHM improved the construction efficiency of multi-parameter driving cycles by 38% under the set conditions.