Balancing risk against utility: Behavior planning using predictive risk maps

摘要: This paper addresses the problem of future behavior evaluation and planning for ADAS in general traffic situations. Complex situations require estimation alternatives terms predictive risks. Based on predicted dynamics scene entities, we present an approach where a continuous, probabilistic model risks is used to build so-called risk maps. These maps indicate how risky certain ego-car trajectory will be at different times so that they can directly plan best possible behavior. Since this optimization highly non-convex combine with sampling-based algorithms RRT∗-type obtain trajectories which minimize maximize utility. We apply our multiple types various scenarios, including inner city highway