Using a Proactive Sensor-System in the Distributed Environment Model Andreas Hermann 1 , Stephan Matzka 2 , and J¨ org Desel 1 1 Department of Applied Computer Science, Catholic University of Eichst¨ att-Ingolstadt, Eichst¨ att, Germany 2 School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, UK [andreas.hermann, joerg.desel]@ku-eichstaett.de, sm217@hw.ac.uk Abstract— In this paper we present the usage of the Dis- tributed Environment Model (DEM) to supply a proactive sensor system. The proactive sensor-system is a novel ap- proach which combines low-resolution range sensors with high-resolution range sensors. Its aim is to achieve a better cost-benefit ratio on getting accurate knowledge about the environment of a car. We describe the main concepts of the DEM and of the proactive sensor system. Furthermore, we show synergistic effects of the DEM environment knowledge base combined with the efficient object classification done by the proactive sensor system. Finally, an evaluation of the DEM and the proactive sensor system in real traffic scenarios is presented. I. I NTRODUCTION In the automobile industry, many comfort and safety ap- plications rely on the knowledge about a car’s environment. Therefore, environmental sensing plays a fundamental role in this field [19]. This holds in particular for Advanced Driver Assistance Systems (ADASs) which gather environmental information via long range radars and cameras in present-day cars. These sensors produce a huge amount of data which has to be transported by the bus system of the car and processed by electronic control units (ECUs). Managing this data is a non-trivial software engineering task. In the AUTOSAR consortium [1], leading automotive manufacturers and suppliers are working together to develop and establish an open industry standard for automotive architectures. Common problems like hardware abstraction and real-time communication with the environment are ad- dressed by this standard. Standard software such as operating systems is specified as well. However, fundamental issues like sensor data distribution and the integration of driving situation awareness [11] are not covered in AUTOSAR. To handle these issues in ADASs development and deployment, we propose the Distributed Environment Model (DEM) as an addition to the AUTOSAR specification. DEM meets the requirements of ADASs by a distributed framework, supporting a situation model, situation analysis and thus situation awareness. The DEM provides a uniform driving situation for all ADASs and thus allows the detachment of situation analysis, resulting in reduced ADASs development complexity. Furthermore, the driving situation is used for internal data distribution in DEM, which leads to a more efficient usage of communication resources. In particular, proactive data distribution for the most likely next driving A. Hermann and S. Matzka are also with Institute for Applied Research, Ingolstadt University of Applied Sciences, 85049 Ingolstadt, Germany situation, as suggested in this paper, improves the perfor- mance of data acquisition from a remote host. However ADASs development heavily relies on an ac- curate knowledge about the car’s environment to detect other traffic participants and possible threats. One way to gain this knowledge is the use of range sensors such as radars or laser scanners. The latter are often capable of acquiring high-resolution range information, yet it is very time-consuming to obtain a regular set of input data. This would for example require the scene to be scanned line by line. In a dynamic road traffic environment this becomes problematic, as a single 3-D scan of the environment takes as long as 4-12 sec [20]. Our objective is to develop a sensor concept, that provides efficient accquisation of high reso- lution environment information. In this context, efficiency stands for an optimum cost-benefit-ratio which is the case if an accurate environmental knowledge can be obtained under real-time constraints. We propose the proactive sensor system to fullfill this requirements. The proactive sensor system aims at increasing sensor resource efficiency by allocating control on high-resolution sensors using a saliency driven utility optimisation scheme (see section III). In this paper we consider the main concepts of DEM and the proactive sensor system and present their synergistic effects. The proactive sensor system profits from information accessible through the DEM. It uses the object trajectories provided by the DEM in order to anticipate the appearance and disappearance of occlusions. As the attempt to observe and classify an occluded object is futile, it decreases sensor resource efficiency. At the same time, the DEM profits from a robust object classification performance because tracking algorithms heavily depend on a good model of object dy- namics, governed by vehicle classes. The remainder of this paper is organized as follows. In section II the DEM architecture is presented. The concept of a proactive sensor system is introduced in section III. Section IV shows evaluation results in real driving scenarios. Summary and conclusion can be found in section V. II. DISTRIBUTED ENVIRONMENT MODEL (DEM) The main concepts of the Distributed Environment Model (DEM) have already been presented in [7]. In this section we give a summary on the major parts of the DEM. The focus is the proactive Sensor System described in section III. 2008 IEEE Intelligent Vehicles Symposium Eindhoven University of Technology Eindhoven, The Netherlands, June 4-6, 2008 978-1-4244-2569-3/08/$20.00 ©2008 IEEE. 703