The latest iteration of Alibaba DAMO Academy's high-precision positioning system, based on a tightly coupled algorithm fused with multiple sensors, achieves centimeter-level positioning that is not dependent on GPS signals.
The system has been deployed in the DAMO Academy's unmanned logistics fleet and is designed through hardware and software integration to achieve industry-leading positioning levels at a cost of 10%.
GPS only provides absolute positioning at the metre level of accuracy, but to increase the accuracy to the centimetre level, the industry's common solution is to use RTK (carrier phase difference technology).
RTKs improve GPS accuracy through ground-based signal enhancement and are often used with high precision inertial guidance. It is costly and does not overcome the problem of unstable or lost satellite signals.
The mainstream multi-source fusion positioning technology is divided into loosely coupled and tightly coupled, the former fuses the sensor data processing results; the latter first fuses the sensor raw data, then centralized calculations, better results and higher difficulty.
DAMO Academy uses a tightly coupled algorithm to realize the fusion of GPS, inertial guidance, wheel speed, camera, LiDAR and other multimodal sensors, using low-precision sensors to achieve centimeter-level positioning, and in tunnels, basements and other places without GPS signals, the positioning system can also operate normally, eliminating the dependence on RTK.
DAMO Academy took this algorithm and implemented it in a soft-hard implementation to develop high precision positioning hardware adapted to the algorithm. The complete positioning system achieves industry-leading high precision positioning at a cost of 10%.
It is worth mentioning that in April this year, the DAMO Academy Autopilot Lab developed a new ISP processor that can improve night vision on autopilot by 10%.
The processor is already used in self-driving logistics vehicles and has industry-leading road test performance.