Occlusion-Robust Method to Determine AGV's Stop Pose by Detecting Reference Object with Template Matching on 2D Environmental Map

We propose a strategy that enables an automated guided vehicle (AGV) to determine its stop position based on the reference objects detected on a 2D environmental map generated by simultaneous localization and mapping (SLAM). Our strategy can save the time and cost required to change the AGV settings after rearranging the room in which the vehicle is used. In industrial applications, AGV stop positions are usually set in relation to a reference object such as a worktable, device, or conveyor used to fulfill their tasks. In our strategy, the reference object template is pre-prepared by cutting the reference object out of a 2D map. If the reference object position changes, the AGV makes a new 2D map of the updated environment, and the template is detected on the new map. After that, the new stop position is calculated based on the detection result. The technical challenge of this strategy is that the shape of the reference object on the map changes when its position and observable sides are changed. Therefore, we propose a template-matching-based detection method that takes occlusion into account. The novel point of our method is the evaluation function for matching, in which the unknown empty areas on the map have different scores to make the detection less sensitive to occlusion. In a verification experiment, we determined that our algorithm could detect a desk with occlusion.