Time-adapted Early Arrival Path for Drone Parcel Delivery through Public Transportation Vehicles: Using Q-learning - Université de Bourgogne Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Time-adapted Early Arrival Path for Drone Parcel Delivery through Public Transportation Vehicles: Using Q-learning

Résumé

Drone-based delivery has gained popularity in recent years, and many different delivery systems and schemes have been proposed. One of the most promising scheme concepts is based on the collaboration between a drone and a public transportation network to expand the delivery range while conserving drone battery energy and reducing delivery costs. Path planning is the main problem with this design, as the public transportation network is stochastic and time-dependent. In this paper, an inspection time-adapted early arrival path problem is formulated, which seeks the path for a drone that ensures: (i) reaching the customer as soon as possible, (ii) adapting to random fluctuations in public transportation schedules, and (iii) taking into acount the battery consumption of the drone. To achieve these requirements, a Q-learning-based planning method is proposed. The simulation results validate the effectiveness and feasibility of Q-learning on the planning path for parcel deliveries: at any departure instant, the arrival of the drones at the customer's location was guaranteed, i.e., the resulting path is 100% reliable. In addition, the convergence of the Q-Learning algorithm was reached after only 1000 learning epochs. Furthermore, the experimental results show that the Q-Learning solution can achieve a lower early arrival time and lower power consumption compared to another algorithm.
Fichier non déposé

Dates et versions

hal-04357344 , version 1 (21-12-2023)

Identifiants

  • HAL Id : hal-04357344 , version 1

Citer

Mohammed Rahmani, Florian Delavernhe, Sidi-Mohammed Senouci, Marion Berbineau. Time-adapted Early Arrival Path for Drone Parcel Delivery through Public Transportation Vehicles: Using Q-learning. ROADEF 2023, Rennes School of Business, en collaboration avec IMT Atlantique et l'Université Rennes 1, Feb 2023, RENNES, France. ⟨hal-04357344⟩
16 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More