Revolutionizing Last-Mile Logistics: A Sustainable Approach Integrating Drones and Trucks

By Samudrapom DamReviewed by Susha Cheriyedath, M.Sc.Jul 9 2023

In an article in the press with the journal Engineering Applications of Artificial Intelligence, researchers proposed a new mathematical model flying sidekick traveling salesman problem integrating deliveries and returns with multiple payloads (FSTSP-DR-MP) to achieve higher sustainability and efficiency in last-mile parcel delivery and return service.

Background

Online retail logistics has become crucial for business operations in recent years owing to the significant growth in online shopping sales, specifically after the outbreak of the coronavirus disease 2019 (COVID-19) pandemic. However, the increase in online purchasing has resulted in more significant carbon dioxide emissions due to the higher utilization of delivery trucks for last-mile delivery. Last-mile delivery accounts for a leading share of overall transportation costs in online sales.

Several studies have demonstrated that the sustainability aspects and operational efficiency of last-mile delivery can be improved using automatic delivery stations, autonomous vehicles and pickup stations, truck-based autonomous robots, and by integrating delivery trucks with couriers on foot and scooters.

Online retail and logistics companies also strive to adopt improved last-mile service methods, such as e-bike and bike couriers, drones, droids, semiautonomous ground vehicles, and autonomous ground vehicles (AGVs). Among these methods, drone-based last-mile delivery has received considerable attention for cost-efficient delivery of small parcels more quickly than conventional truck-based delivery methods. Moreover, drones are also eco-friendly as they produce substantially lesser amounts of carbon during operation compared to trucks.

Recently, delivery systems based on both trucks and drones have been developed to improve the sustainability and efficiency of last-mile delivery. However, all truck and drone-based delivery approaches have only focused on parcel deliveries while ignoring parcel return services, which have increased significantly with the expansion of e-commerce.

Currently, online consumers return 15-40% of ordered goods in conventional ways, such as by dropping them off at designated locations or sending them through post offices, necessitating additional traveling and contributing to carbon dioxide emissions.

New model for sustainable last-mile parcel return and delivery service

In this paper, researchers proposed a new mathematical model FSTSP-DR-MP for integrating both returns and delivery service in a combined drone and truck operation. They formulated the problem as a mixed-integer linear program (MILP) to minimize the total service time of the model.

The FSTSP-DR-MP model is a nondeterministic polynomial (NP)-complex problem derived from the standard TSP and FSTSP models. Researchers compared their proposed model with TSP and FSTSP models to determine the improvement in total truck time and total service time savings while considering multiple return customers and payloads. Valid inequalities were introduced in the model to reduce the mathematical model runtime significantly.

In this model, a truck and a drone were coordinated to perform delivery and return/pickup services in a single drone sortie while including several payloads. Parcels were delivered to a customer either by a drone launched from the truck or by the truck, while returns were collected from customers by drone or by truck.

The drone performed multiple stops, a new feature in any truck and drone-based last-mile delivery models consistent with the ongoing research on developing drones with greater endurance and weight/payload capacity. This feature allowed the exploitation of the differences between return and delivery customers.

Small-size cases were solved using the MILP implemented in CPLEX Python API software. However, researchers proposed a meta-heuristic derived from variable neighborhood search (VNS) that can iteratively build drone and truck routes to solve realistic size cases.

Solutions with only a 4.8% mean optimality gap on 10 customer instances can be obtained through this meta-heuristic. Moreover, it was not extremely sensitive to randomness and quickly handled realistic problem sizes of up to 100 customers. The FSTSP-DR-MP model was also compared with the multi-visit traveling salesman problem with multi-drones (MTSP-MD) to evaluate the performance of the single drone with multiple payloads and multiple drones with a single payload.

The critical difference between the proposed FSTSP-DR-MP model and the standard FSTSP model is the highest number of possible stops per drone sortie. In the standard FSTSP model, the maximum number of stops for each drone sortie is two customers as the model cannot differentiate between delivery and pickup.

However, the proposed model can make a maximum number of four stops, which results in more significant savings in potential service time compared to a standard FSTSP model, improved drone utilization, and increased customer expectations concerning return services, leading to higher sustainability of the model.

Significance of the study and future outlook

Researchers demonstrated that the FSTSP-DR-MP model can work effectively for up to 100 customers, the usual number in last-mile logistics. The number of stops of each drone sortie was increased by exploiting the differences in delivery and return customers to reduce total truck travel time and service time.

The integration of delivery and return truck-drone operations decreased the total truck travel time significantly, which led to a 22.9% and 36.2% reduction in diesel engine emissions when compared to FSTSP/one truck and one drone-based delivery schemes and TSP/traditional single-truck based delivery schemes, respectively, based on 10 customer instances.

Moreover, the overall service times were reduced by 18.3% and 32.5% compared to FSTSP and TSP, respectively. A single drone with multiple payloads outperformed multiple drones with a single payload with an average 3.9% improvement in total service time.

To summarize, the findings of this study demonstrated the feasibility of using this proposed FSTSP-DR-MP model for sustainable last-mile logistics. However, more research is required to improve the total service time savings compared to FSTSP and TSP by considering multiple drones and trucks in the last-mile delivery.

Drone scheduling management will be necessary while using multiple drones to avoid congestion or mid-air collision. Additionally, drones and trucks with varying speeds can be used to conserve battery life or fuel and maintain/improve service time savings.