Intention Recognition in Depots with Automated Guided Vehicles

Whenever automated guided vehicles (AGVs) and humans work in a common environment, dangerous situations can emerge, especially if the human workers do not know or do not pay attention to the current activities of the automated vehicles. A complete emergency stop in all situations where a worker could interfere with a vehicles plan would not be a good solution due to the associated costs. In this article, we propose an approach which is based on a qualitative scene representation created from sensory input. By detecting actions in the scene, hypotheses about possible plans are generated. Taking into account possible future actions of workers, a risk assessment is performed. The behavior of an automated vehicle has to be changed only in those situations that exceed a certain risk threshold.