Decision-Making in Automotive Collision Avoidance Systems
This thesis is concerned with decision-making in systems that can detect hazardous traffic situations and assist drivers in avoiding collisions by using automatic braking or steering. The aim with these systems is to reduce the number of accidents and their consequences without disturbing the driver with unnecessary interventions during normal traffic conditions.
The main contribution of this thesis consists of algorithms for evaluating if the driver needs assistance to avoid colliding with a single road user in any traffic situation. The proposed algorithms, which are shown to work well in a real-time environment, are evaluated using data from both real traffic conditions, simulations and collision situations on a test track. Moreover, a probabilistic decision-making framework is presented for jointly evaluating the driver acceptance of an intervention and the necessity thereof to automatically avoid an accident. The framework enables earlier interventions in critical traffic situations, thereby increasing the benefit of the system. Additionally, a method is proposed for estimating driver distraction by observing the driver's steering behavior prior to near-crash situations. It is shown that earlier interventions can be triggered when the driver is assessed as being distracted without significantly increasing the risk of unnecessary interventions. Decision-making on when to assist the driver by steering and when to assist by braking is discussed and an algorithm for finding suitable evasive steering maneuvers to pass between multiple moving objects is presented.