Intelligent Scheduling and Charging Optimization for CAEVs and UAVs in Urban Transportation Networks

Abstract

Urban transportation systems are increasingly strained by congestion, air pollution, and greenhouse gas emissions, motivating the need for scalable and sustainable mobility solutions. Connected Autonomous Electric Vehicles (CAEVs) and Uncrewed Aerial Vehicles (UAVs) are key enablers of next generation intelligent transportation systems (ITS), yet their adoption remains constrained by limited battery capacity, fragmented charging infrastructure, and the absence of reliable coordination mechanisms under dynamic and heterogeneous conditions. Existing research often addresses individual charging technologies or isolated vehicle classes, leaving a gap in system-level coordination across ground and aerial platforms. This thesis addresses this gap by demonstrating that charging coordination, rather than charging technology alone, is a critical barrier to sustainable deployment of autonomous electric mobility. A scalable and interoperable three-layer charging network architecture is proposed to support static charging, dynamic wireless charging, and vehicle-to-vehicle (V2V) charging across heterogeneous charging node types. A hybrid coordination topology and unified handshake protocol enable interoperable charging discovery, reservation based scheduling, and execution while remaining compatible with existing communication infrastructure. Building on this architectural foundation, coordinated charging scheduling and trip planning mechanisms that consider both vehicle routing and charging resource allocation are developed using heuristic and learning-based approaches. Two heuristic strategies, Static Heuristic Charging Scheduling Policy (SH-CSP) and Dynamic Heuristic Charging Scheduling Policy (DH-CSP), provide interpretable baselines and handle early and late arrivals. To address scalability and non-stationary demand, Safety, Sustainability, and Scheduling-Aware Feasibility Enhanced Deep Deterministic Policy Gradient (SAFE-DDPG) is introduced to enable adaptive, real-time scheduling under heterogeneous charging conditions. Reliability and fairness are embedded as system-level properties within the architecture and scheduling framework. Simulation-based evaluation demonstrates charging reservation fulfillment rates approaching 98%, waiting time reductions of up to approximately 80% relative to heuristic baselines, and consistently high fairness across heterogeneous charging networks. Comparative evaluation against baseline deep reinforcement learning methods, including DDPG and TD3, and explainable artificial intelligence (XAI) techniques further validate effectiveness and interpretability. Overall, this thesis presents a unified, reliable, and scalable framework for intelligent scheduling and charging optimization through learning-based decision-making in next-generation urban transportation systems.