Collision warning method for workshops based on dynamic target trajectory prediction

To address collision risks between autonomous guided vehicles and pedestrians in workshop environments, this study proposes a trajectory prediction-based collision warning method for unmanned forklifts and pedestrians. A pedestrian trajectory prediction model based on an autoencoder is first designed to enable end-to-end trajectory prediction using factory surveillance data. The three-dimensional human-machine distance is then computed based on monocular camera distance recovery principles. A fuzzy logic-based collision risk grading and warning model is developed to quantify potential collision risks into actionable warning levels for real-time safety interventions. Experimental results demonstrate that the proposed method effectively detects and predicts pedestrian movements in dynamic workshop environments with high confidence and low latency, reducing potential collision risks and improving workplace safety.