Drivers' Manoeuvre Prediction for Safe HRI

摘要: Machines with high levels of autonomy such as robots and our growing need to interact them creates challenges ensure safe operation. The recent interest create autonomous vehicles through the integration control decision-making systems makes too. We therefore applied estimation mechanisms currently investigated for human-robot interaction human-vehicle interaction. In other words, we define vehicle an agent which human driver interacts, focus on understanding intentions processes. These are then integrated into ro-bot‘s/vehicle's own system not only understand behaviour while it occurs but predict next actions. To obtain knowledge about human's intentions, this work relies heavily use motion tracking data (i.e. skeletal tracking, body posture)gathered from drivers whilst driving. a data-driven approach both classify current driving manoeuvres future manoeuvres, by using fixed prediction window augmenting standard set manoeuvres. Results validated against different sizes, seat preferences expertise evaluate robustness methods; precision recall metrics higher than 95% manoeuvre classification 90% time-windows up 1.3 seconds obtained. idea adds highly novel aspect human-robot/human-vehicle interaction, allowing decision at later point.