An observation model for high resolution radar data in the context of an automotive pedestrian safety system

Within the EU-Project ARTRAC a radar sensor is developed that offers new capabilities for enhanced pedestrian protection. In this paper we describe a sensor model for high resolution measurements as a first step to help improving the object detection and tracking. For the tracking, with the commonly used Kalman filter and its derivations, one makes several assumptions which do not necessarily hold for pedestrian tracking. Therefore we propose to use a non-linear filter e.g. the Particle filter. In order to define an observation model and gain a suitable proposal distribution, we make use of grid structures to represent parts of the measurement and system state space. Also, we set several assumptions to reduce the computational costs of the tracking process.