SLAM system for mobile robots using fused vision and motion sensor data

This technology is a SLAM system for mobile robots that improves localization precision by combining environmental feature points and spatial occupancy information extracted from vision sensors with movement data calculated from motion sensors using a probability-based data fusion technique.

When using vision sensors alone, issues such as motion blur and decreased accuracy in dynamic environments often led to cumulative errors in localization and mapping.

This technology proposes a modular SLAM architecture that includes a vision sensor processor, a motion sensor processor, and a third processor that fuses this information using probability-based filters, such as a Kalman filter, to re-estimate the robot's position. This approach suppresses error accumulation through sensor complementarity, enabling the creation of precise maps. It can be applied to various indoor autonomous driving applications, including cleaning robots, logistics robots, and service robots, significantly enhancing the reliability of localization by mitigating error accumulation through sensor fusion.

‍

‍

Key Features:

• Vision sensor processor that extracts environmental feature points and spatial occupancy information from vision sensors such as cameras or lasers
• Motion sensor processor that calculates movement information from motion sensors such as accelerometers, gyroscopes, and encoders
• Third processor that fuses information from both processors using probability-based filters such as a Kalman filter
• Modular SLAM architecture that re-estimates the robot's position and generates maps based on the fused information