Data stream management systems (DSMSs) are suitable for managing and processing continuous data at high input rates with low latency. For advanced driver assistance including autonomous driving, embedded systems use a variety of onboard sensor data with communications from outside the vehicle. Thus, the software developed for such systems must be able to handle large volumes of data and complex processing. We develop a platform that integrates and manages data in an automotive embedded system using a DSMS. However, because automotive data processing, which is distributed in in-vehicle networks of the embedded system, is time-critical and must be reliable to reduce sensor noise, it is difficult to identify conventional DSMSs that meet these requirements. To address these new challenges, we develop an automotive embedded DSMS (AEDSMS). This AEDSMS precompiles high-level queries into executable query plans when designing automotive systems that demand time-criticality. Data stream processing is distributed in in-vehicle networks appropriately, where real-time scheduling and senor data fusion are also applied to meet deadlines and enhance the reliability of sensor data. The main contributions of this paper are as follows: (1) we establish a clear understanding of the challenges faced when introducing DSMSs into the automotive field; (2) we propose an AEDSMS to tackle these challenges; and (3) we evaluate the AEDSMS during run-time for advanced driver assistance.
2017 by the Information Processing Society of Japan