Machine Learning-Based Object Movement Prediction Method Using Occupancy Grid Maps from Roadside Sensor

For automated vehicles, wide-ranging and real-time detection of the surrounding environment and accurate recognition of objects, including pedestrians, vehicles, and their movements, are crucial. In previous work, we proposed a method for estimating road environments as an occupancy grid map (OGM) using roadside sensors. However, OGMs independently calculate occupancy probabilities for each cell, which poses a challenge in accounting for the movement of objects across cells. This study proposed a machine learning-based method for predicting future OGMs, using OGMs from roadside LiDAR sensors. Real-world evaluations demonstrated that the proposed method predicts object movement with short execution times.