Abstract
In the rapidly evolving landscape of transportation, mobility, and logistics, the last-mile represents the final and crucial leg of the delivery journey. It involves goods travelling from a transportation hub to the ultimate destination. This is the most expensive and time-sensitive part of the supply chain business model. Challenges include navigating dense urban environments with various types of traffic participants (e.g., pedestrians, bicycles, animals, electric scooters, motorcycles, etc.), dealing with traffic congestion, locating specific delivery points, managing a high density of stops, and handling failed delivery attempts. In this context, the intersection of edge artificial intelligence (AI), autonomous systems, robotics, and sensor fusion in perception and navigation advances the development of last-mile delivery autonomous vehicle (AV) platforms that evolve towards software-defined and AI-defined vehicles (SDVs and ADVs). The advancements include multiple sensor systems for perception and communication (e.g., ultrasound, inertial, LiDAR, radar, camera, V2X, etc.), real-time data processing for localisation, and robust algorithms for navigation and interaction with diverse traffic environments. This chapter presents the concept and the implementation of an AI-based perception and sensor fusion platform technical solution for autonomous lastmile delivery in controlled traffic environments.