RoboSense RS-Reference For Validation Of Perception Systems

Testing and validation of perception system is very important for the safety of autonomous driving applications, out of which, autonomous driving requires large amounts of road tests

However, obtaining massive and accurate ground truth data is a challenge. Manual data labelling requires extremely long time and huge cost. Needless to say, there is also the challenge of lack of efficient tools, standards and customized analysis for validation of different kinds of sensors.

RoboSense combines high-performance sensor systems with mature AI perception algorithms to provide a set of accurate, efficient, complete, and customizable sensing performance uation tool chains. The system not only solves the problems of high-cost, low-efficiency and error-prone manual calibration, but also makes up for the lack of uation tools and immature tool chains, which can greatly improve the efficiency of project development and accelerate the mass production process of Autonomous Driving and ADAS applications.

the ground truth output by the RS-Reference 2.0, and the real-time road image

The RS-Reference Features:

 High-Performance and Mature Sensor System

The RS-Reference can fuse and utilize the respective advantages of various high-performance and mature sensors including LiDAR, millimeter-wave radar, RTK, and camera to improve the accuracy of ground truth to a large extent.

 Powerful Offline Processing Software Algorithms

The RS-Reference provides customized offline sensing algorithms, which is developed based on the 12 years’ experience of research and development of RoboSense LiDAR AI perception algorithms. The offline algorithm utilizes a variety of high-performance algorithms technology for offline optimization of objects detection and tracking with sufficient computing resources to obtain accurate and reliable ground truth data, breaking the limitation of real-time processing algorithms.

 Full-Stack Toolchains

The RS-Reference provides a complete full-stack toolchain including tools of collection, visualization, manual verification and uation.

 Customized Test For Diverse Sensors

The RS-Reference provides dedicated uation modules based on the acteristics of different types of sensors including LiDAR, millimeter-wave radar, cameras etc, and can be customized.

 Integrated Installation Design

The RS-Reference adopts an integrated roof bracket to install the sensors, which can be deployed quickly without any car body modification and can directly uates ground truth data of the autonomous driving perception system.

The RS-Reference can assist OEM, Tier1 and research institutes, provider of millimeter-wave radar and cameras, etc to perform highly efficient test and validation and accelerates mass production of ADAS and AD applications.