An enhanced adaptive large neighborhood search for fatigue-conscious electric vehicle routing and scheduling problem considering driver heterogeneity

Driving fatigue is an important safety hazard for logistics enterprises as it increases accident risks and deteriorates drivers' well-being. To effectively manage driving fatigue, a novel problem structure named fatigue-conscious electric vehicle routing and scheduling problem considering driver heterogeneity (FEVRSPH) is investigated in this paper. To address both profit and fatigue concerns, hierarchical objectives are proposed in the model of FEVRSPH, with minimizing the number of vehicles as the primary objective and minimizing the maximum driving fatigue as the secondary. In particular, the impact of driver heterogeneity in terms of circadian rhythm and driving time on fatigue accumulation and alleviation is considered in the model. To address the computational challenges of FEVRSPH, an enhanced adaptive large neighborhood search (EALNS) is introduced in this paper. Several enhancement strategies based on the problem knowledge are designed in EALNS, including an independent search mechanism of customer nodes enabled by the feasibility test operator to enlarge the search space, a modified initialization method to improve the quality of initial solutions, and a reverse scheduling approach to rapidly reduce the maximum driving fatigue. The following findings are obtained from the comprehensive experiments: (1) the proposed strategies considerably enhance the performance of EALNS; compared with four state-of-art meta-heuristics, EALNS reduces the average values of the number of vehicles by 3.63%-10.61% and obtains the optimal average fatigue index in 11 out of the 12 sets; (2) considering driver heterogeneity is necessary for practical applications as it improves the quality of the solution and avoids violating time windows and fatigue thresholds; (3) by directly optimizing driving fatigue, the average fatigue index is reduced by 14.97%-20.58% at the cost of a maximum 6.09% increase in driving distance; (4) the proposed model and EALNS are effective in real-world applications by evaluating a case study.