Method for generating an intersection pattern recognition model using sensor data from a mobile robot and an intersection pattern recognition system

Method for Generating Intersection Pattern Recognition Models Using Mobile Robot Sensor Data and Intersection Pattern Recognition System

This technology is a control mechanism that recognizes intersection patterns and determines the absolute position of a mobile robot by converting pre-established guide lines or structural data into a grid-based path map and training a normalized virtual map alongside sensor data, without the need for additional landmark installations.

Conventional technologies require landmark sensors at every intersection to estimate a mobile robot's position, leading to high installation costs, a lack of flexibility when environmental changes necessitate reinstallation, and difficulties in maintaining real-time performance due to increased data transmission volume when operating multiple robots.

This technology generates a path map based on usage environment information within a management server, creates a virtual map normalized so that the distance between intersections is an integer multiple of a unit length, and provides the robot with a model that classifies and recognizes 'L', 'T', and '+' shaped intersection patterns by training this data with sensor data. It can be applied to logistics transport, service robots, and autonomous driving platforms, enabling real-time movement control of mobile robots without separate landmark sensors, thereby improving the efficiency and cost-effectiveness of such physical distribution systems.

‍

‍

Key Features:

• A training unit that generates an intersection pattern recognition model using virtual maps and mobile robot sensor data as training data.
• A map generation unit where the management server receives usage environment information for the mobile robot and generates a path map based on that information.
• A normalization unit that generates a virtual map by normalizing the path map according to pre-set rules.
• An intersection pattern recognition system using mobile robot sensor data, wherein the path map consists of a grid-based map, and the normalization unit generates the virtual map by normalizing the path map so that the distance between intersections is an integer multiple of a unit length.

‍

This invention was developed with the support of the Ministry of Science and ICT's AI-based Anti-Drone Active Control Technology Development project.