In the field of autonomous driving, low-speed scenarios (with a speed of ≤30km/h) are often regarded as the "simplified version of technology", but it is precisely this track that has given birth to the world's first commercialized unmanned driving project - the Amazon Kiva robot. From the magnetic stripe navigation of AGVs (Automated Guided Vehicles) in the 1950s to the "six-dimensional perception" system integrating 5G, AI and V2X in 2025, the evolution history of low-speed unmanned driving is actually an industrial revolution where scenarios define technologies and demands drive innovation.
Technology Generation: From "Ground Stickers" to "All-Domain Perception"
The first generation: Magnetic stripe/QR Code Navigation (1950s-2010s)
Technical features: It relies on the preset path on the ground and is located by magnetic nails or QR codes. A typical case is the early factory AGV.
Limitations: Path solidification and high expansion costs, only suitable for highly structured scenarios.
Second Generation: Laser SLAM Navigation (2010s-2020s)
Technological breakthrough: LiDAR combined with inertial navigation achieves "mark-free positioning", with the representative product being the Geek+ warehouse robot.
Application expansion: Extending from factories to e-commerce warehouses, but the dynamic obstacle avoidance capability is limited and virtual boundary Settings are still required.
The third generation: Multi-Sensor Fusion +AI Decision-making (20S - Present)
Technology upgrade: LiDAR + vision +IMU multi-modal perception, combined with deep learning algorithms, to achieve real-time path planning in complex scenarios.
Typical case: JD Logistics' "Sirius" unmanned vehicle can make autonomous decisions in a mixed pedestrian and vehicle park, and the obstacle avoidance response time is compressed to 50ms.
The Fourth Generation: Vehicle-Road-Cloud Collaboration (2025+)
Technological revolution: Real-time interaction with roadside devices and cloud brains through 5G networks to obtain beyond-visual-range perception information.
Scene breakthrough: The unmanned container trucks at Shenzhen Mawan Port can be dispatched through a cloud platform to achieve safe mixed driving with manned vehicles.
Scene Revolution: From "Replacing Human Labor" to "Reconstructing Processes"
E-commerce Warehousing: The "First Battlefield" for Efficiency Enhancement
Pain point: Traditional warehouse manual picking is inefficient and has a high error rate.
Solution: The Geek + PopPick system achieves "goods-to-person" picking through dynamic storage location optimization and robot cluster scheduling, increasing efficiency by 300% and achieving an accuracy rate of 99.99%.
Hospital Logistics: The "Lifeline" for Precise Distribution
Pain point: The transportation of drugs and specimens within hospitals requires strict time and temperature control.
Solution: The Noah Hospital logistics robot is equipped with an ultraviolet disinfection module and temperature and humidity sensors. It can independently plan the use of elevators, achieving a direct connection from the operating room to the blood bank within 5 minutes.
Industrial scenarios: The "nerve endings" of flexible Manufacturing
Pain point: Traditional AGVs are difficult to adapt to the production mode of small batches and multiple varieties.
Solution: SEER's AMR (Autonomous Mobile Robot) supports natural navigation and modular task execution, enabling rapid production line switching. The deployment time has been reduced from 3 days to 3 hours.
Public Services: "Smart Connection" for the Last Mile
Pain point: The demand for short-distance transportation in scenic spots and industrial parks is fragmented.
Solution: Baidu Apollo's "Neolithic" unmanned retail vehicle, integrating L4-level autonomous driving and intelligent retail systems, can dynamically plan routes in scenic spots and serve over 500 people per day.
Challenges and the Future: The Triple Leap from "Usable" to "User-friendly"
Technical Bottleneck: The "Last Centimeter" in Complex Scenarios
Dynamic mixed traffic of pedestrians and vehicles: The behavior of pedestrians within the park is unpredictable, and it is necessary to strengthen the training of social behavior models.
Extreme weather adaptation: The performance of lidar deteriorates in rainy and snowy weather, and it is necessary to integrate millimeter-wave radar and visual redundancy solutions.
Business model: From "Equipment sales" to "Service subscription"
Equipment manufacturer transformation: Hikrobot has launched the "Robot as a Service (RaaS)" model, charging based on usage to reduce the initial investment for customers.
Data value mining: By optimizing the scheduling algorithm through accumulated scene data, a closed loop of "data - algorithm - service" is formed.
Standards and Regulations: The Art of Balancing Safety and Efficiency
Functional safety certification: According to the ISO 21448 (SOTIF) standard requirements, it needs to be verified through both virtual simulation and real circuit testing.
Responsibility Definition: Shenzhen has issued the "Road Testing Specifications for Low-speed Unmanned Vehicles", clearly defining the principles for accident responsibility division.
From AGVs to vehicle-road-cloud collaboration, the evolution of low-speed unmanned driving has never stopped. It is not merely an iteration of technology, but also an ultimate pursuit of "efficiency, security and flexibility". When warehouse robots become the "invisible arteries" of e-commerce, when hospital logistics vehicles transform into "life ferrymen", and when industrial AMRs restructure the underlying logic of flexible manufacturing, what we see is a future where technology redefines the value of scenarios. This revolution may not have the grandeur of L4-level autonomous driving, but it has already reshaped the texture of industries and life in silence.
