This technology maps FMCW radar target detection data into a 2D angle-velocity domain and calculates a robot's movement speed and rotation angle without external sensors by analyzing data correlations between consecutive scans and applying linear regression to trend lines, enabling simultaneous localization and mapping (SLAM) through ego-motion estimation.
Conventional SLAM systems require additional hardware such as motor encoders or gyro sensors to estimate a robot's ego-motion, which increases system complexity and cost. Furthermore, laser and camera-based sensors often suffer from performance degradation in low-light or adverse weather conditions.
This technology proposes a method that converts the relative velocity and angle information of targets acquired from radar sensors into a binary matrix, derives the rotation angle through forward and backward cross-correlation operations, and calculates movement speed from the velocity-axis intercept by performing linear regression on the trend line of detection points in the angle-velocity domain. This allows for robust localization in adverse conditions without the need for additional sensors. Because it functions even in environments where cameras and LiDAR are ineffective—such as smoke, dust, or low-light conditions—it significantly enhances the reliability of disaster response robots and industrial autonomous equipment.
This invention was developed with the support of the Ministry of Science and ICT's "Research on Next-Generation Radar Systems Robust to Interference: Deep Learning-Based Data Augmentation Core Technology and Adaptive Target Tracking Technology."
N/A
