JD Autonomous Delivery Robots: Technologies, Patents, and Future Challenges

JD's delivery robots have become a familiar sight on campuses and industrial parks, and on June 18 they achieved the world's first full‑scenario, all‑weather, regularized delivery operation, covering residential, campus, and courier‑hand‑off scenarios with a new multi‑vehicle, multi‑route scheduling model.

The current third‑generation robots integrate advanced technologies developed by JD's autonomous driving team, including a multi‑sensor fusion system that combines inertial navigation, LiDAR point‑cloud matching, visual positioning, and wheel‑speed data to achieve centimeter‑level localization accuracy.

To navigate complex campus and park environments, the robots employ large‑scale data collection, point‑cloud clustering, and deep‑learning‑based radar perception, delivering robust tracking at over 50 Hz even under heavy pedestrian and vehicle flow.

Real‑time local mapping is created through SLAM techniques, merging radar and visual information to enhance environmental awareness and support reliable path planning.

Deep learning and reinforcement learning are applied to motion control, using sensor inputs to continuously refine trajectory curvature, deviation, and heading, enabling the robots to recognize traffic signs and lane markings under varying lighting and weather conditions, and to adjust speed and posture smoothly from low to medium speeds.

JD has filed hundreds of patents covering vehicle structure, electrical systems, perception modules, algorithms (sensor fusion, mapping, deep learning, control), and safety features, protecting its innovations across hardware and software.

Future challenges include the need for high‑precision maps, which are costly to acquire and produce, and the high price of LiDAR sensors, both of which limit large‑scale deployment.

The article concludes with a call for public support in the "China Good Patent" voting campaign, encouraging readers to vote for JD's delivery robot patents via WeChat, mobile web, or desktop web links.