Autonomous Last-mile Delivery Vehicles in Complex Traffic Environments

Abstract: E-commerce has evolved with the digital technology revolution over the years. Last-mile logistics service contributes a significant part of the e-commerce experience. In contrast to the traditional last-mile logistics services, smart logistics service with autonomous driving technologies provides a promising solution to reduce the delivery cost and to improve efficiency. However, the traffic conditions in complex traffic environments, such as those in China, are more challenging compared to those in well-developed countries. Many types of moving objects (such as pedestrians, bicycles, electric bicycles, and motorcycles, etc.) share the road with autonomous vehicles, and their behaviors are not easy to track and predict. This paper introduces a technical solution from JD.com, a leading E-commerce company in China, to the autonomous last-mile delivery in complex traffic environments. Concretely, the methodologies in each module of our autonomous vehicles are presented, together with safety guarantee strategies. Up to this point, JD.com has deployed more than 300 self-driving vehicles for trial operations in tens of provinces of China, with an accumulated 715,819 miles and up to millions of on-road testing hours.

• The paper demonstrates a comprehensive, modular approach with over 20 key modules for autonomous navigation in urban delivery settings.
• It employs robust techniques that fuse GPS, Lidar, and visual data to achieve precise localization and reliable object detection amid unpredictable traffic.
• Results indicate potential savings of up to $110 million annually, underscoring significant enhancements in operational efficiency and safety.

Autonomous Last-mile Delivery Vehicles in Complex Traffic Environments

This paper addresses the integration of autonomous last-mile delivery vehicles in intricate traffic conditions, particularly within the context of urban China. E-commerce companies, such as JD.com, face significant logistic challenges in these environments due to the presence of various types of road users who often do not adhere to traffic regulations. The aim is to reduce delivery costs, enhance efficiency, and overcome the limitations posed by traditional last-mile logistics services.

Context and Motivation

The paper outlines the motivation for employing autonomous vehicles in urban last-mile delivery, emphasizing cost reduction and improved operational efficiency. With the need to reduce reliance on labor due to escalating costs and inefficiencies associated with human delivery processes, autonomous driving technologies present a viable solution. The challenge lies in navigating China's dense urban centers, characterized by high residents living in apartments and diverse, often unpredictable road participants, including bicycles and pedestrians.

Technical Solutions

JD.com's approach incorporates a comprehensive architecture involving over 20 modules crucial for autonomous operation. This modular system combines both online and offline functionalities, ranging from real-time localization and perception to decision-making and planning.

• Localization and HD Mapping: The localization solution relies on multiple data sources, including GPS, Lidar, and visual odometry, synthesized through algorithms such as Iterative Closest Point (ICP) within a Kalman filter framework. The HD maps are developed to reflect the dynamic road conditions in urban Chinese settings, requiring frequent updates and cost-effective construction methods.
• Perception Systems: Object detection within the perception module employs machine-learning and geometry-based methods to recognize and track various road dynamics. The fusion of lidar and camera data plays a crucial role here, addressing challenges such as pedestrian recognition.
• Prediction and Planning: A two-layered prediction model estimates future trajectories of dynamic traffic components. Planning modules execute trajectory decisions via quadratic programming, optimizing for safety and efficiency amidst complex traffic interactions.

Safety and Monitoring

Safety strategies are meticulously woven into vehicle architecture through multiple layers, including simulation-based verification, vehicle-end monitoring, and a robust remote monitoring platform. These ensure operational reliability and enhance the decision-making framework, especially during unforeseen events.

Implications and Future Directions

The implications of deploying autonomous delivery vehicles are substantial. The economic impact is underscored by predicted cost savings of up to $110 million annually by incorporating these technologies into 10% of JD.com's delivery processes. Future advancements will likely focus on optimizing decision algorithms and enhancing the robustness of perception systems to handle even more complex urban scenarios.

As autonomous vehicle technology continues to mature, deployment will gradually shift toward handling higher-speed operations with increasing autonomy. Clear delineation between machine-handled tasks and human oversight will maximize both safety and efficiency. Moreover, the lessons learned regarding clear task separation between human and robotic operations highlight the evolutionary path in the autonomous delivery landscape.

In conclusion, this paper provides a detailed exploration of the technical and practical considerations for implementing autonomous last-mile delivery systems in complex urban environments. The research presents a framework that, while initially targeted for challenging Chinese metros, may offer insights applicable to broader global contexts as cities grow denser and e-commerce demands continue to rise.