In the logistics industry, the use of electric vehicles (EVs) in delivery operations is required to achieve a sustainable society. Since EVs have a shorter driving range than internal combustion engine vehicles (ICEVs), charging between deliveries is needed to extend the driving range of EVs. However, since the budget of the initial investment for electrifications is limited, the number of chargers installed at logistic depots may be less than the number of EVs. In addition, there may be fewer parking lots than EVs at logistics depots. For these reasons, it is difficult to charge all EVs simultaneously. Therefore, an EV fleet charging management solution is needed to replace ICEVs with as many EVs as possible.
This research focuses on logistic depots, where vehicles are responsible for multiple delivery jobs each day, and a parking lot with a limited number of chargers is facilitated. When a delivery job is completed, the vehicles return to the depot and, if necessary, are recharged before the next delivery job. All delivery jobs are given in advance and must be assigned to vehicles. The first challenge of managing EV fleet at such a depot is that a plan that incorporates both delivery and charging is necessary. The second challenge is the power shortages caused by the delay in vehicle’s arrival. This is because, if an EV arrives at the depot later than scheduled, its charging time may be shortened, which may lead to power shortages in subsequent delivery jobs. To address this issue, a robust EV fleet management method which accommodates uncertain arrival times is needed. The third challenge is the computation time limitation. We aim to solve this problem in a short time to apply this method to on-site operations.
To overcome the challenges, we developed a robust EV fleet management method that integrates the determination of delivery job assignment, charging plan and parking plan for EVs, while considering the accommodation of robustness against uncertain delays. To solve this problem within a limited time, we propose a method that divides the problem into two steps: (1) generating diverse solutions and then, (2) evaluating their robustness under uncertainties. At step (1), we applied several parameterized constraints that may have an impact on power shortages, and help solutions be more resilient in terms of power, charging time, and interval time.
Computational experiments show that a constraint on the lower bounds of possible charging time has the most impact on the result, and our method can replace 94\% of ICEVs with EVs, an increase of 9 percentage points over a conventional method that does not consider uncertainties.

Ultra-fast delivery revolutionizes food and grocery services, with several companies advertising delivery times under 15 to 30 minutes. Motivated by the multi-billion-dollar industry that has emerged in recent years within the delivery business, we investigate the network design problem for ultra-fast delivery services. This involves decisions on micro-depot locations and customer allocations, considering various service guarantee levels. We develop robust probabilistic envelope constrained (PEC) programs to handle uncertainties in travel times and customer order arrivals, and jointly optimize the protection level to avoid both excessive risk and conservatism. To enhance the tractability of PEC models, we derive their equivalent semi-infinite linear programs and propose inner and outer approximations with finite linear constraints. We validate the accuracy of these approximations through extensive experiments using real-world data from Amazon and the Google API, along with a comparative study of different formulations. Varying service levels in ultra-fast delivery affect profitability and reliability, contingent on service level definitions and compliance probabilities of these guaranteed service levels. We find that a daily service level with multi-layer partial protection outperforms other policies investigated in this paper, yielding higher profitability and mild violations of service level guarantees, and it proves to be an effective strategy for profitable and reliable ultra-fast delivery without over-committing or under-delivering, regardless of ordering times or traffic conditions. Additionally, empirical evidence indicates that providing ultra-fast delivery in rural areas poses unique challenges compared to urban settings.

This study introduces an advanced mathematical optimization approach for efficient scheduling and routing of technicians in field service operations, aiming to maximize profitability and customer satisfaction while addressing uncertainty. It categorizes customers into mandatory and optional visits and assigns technicians to vehicles based on skill levels and operational constraints. The model incorporates stochastic elements to manage unpredictability in travel and service times. The uncertainty in travel and service time is modeled using chance constraints. Constraints guarantee precise technician-to-vehicle assignments, feasible routes, adherence to time limits, and appropriate skill levels for each service. The problem is solved using a logic-based Benders decomposition approach that decomposes the original problem into a master problem and a series of subproblems.

We present a contextual framework for solving the data-driven traveling salesman problem in last-mile delivery. The objective of the framework is to generate routes similar to historical high-quality ones, as classified by operational experts, by considering the unstructured and complex features of last-mile delivery operations. The framework involves learning a transition likelihood matrix between customer stops and using a classical TSP solver to generate routes similar to those high-quality routes in a given dataset. We employ feature-based factorization of the transition likelihood matrix, which reduces the dimensions of the information to be learned. We develop a derivative-free algorithm by extending the coordinate search method and a complementary method for learning the transitions directly from the data. We test the efficiency of the methods using a case study based on the Amazon Last-Mile Routing Challenge and show that our methods are effective, interpretable, and flexible. A preliminary version of our methods achieved third place in the aforementioned challenge. We improve on this result by 23\% in an out-of-sample dataset.