In English

Precise time-stamping and delays measurement in embedded sensor network

Georgia-Eirini Diakou
Göteborg : Chalmers tekniska högskola, 2016. 60 s.
[Examensarbete på avancerad nivå]

The field of active safety in the automotive industry has unlimited potentials for the future development of vehicles, especially in autonomy, which is expected to change the way we drive in near future. In recent years, safer cars have become one of the top-priorities among the biggest car manufacturers. The driver support functions that make cars safer rely on accurate onboard sensor readings of the surroundings of the car. To assure the high performance of the active safety systems as well as to avoid problems for their verification during the development stage, it is necessary for the support functions and the sensors to have the same definition of time when an obstacle is detected on the vehicle’s surroundings. This requires synchronizing the sensors’ data with data from other nodes where sensor fusion and decision making are run for example. Poor synchronization directly inserts unwanted delays between the nodes and impacts the ability of the functions to intervene on time, causing serious safety issues. The present thesis addresses the problem of existence of delays in a real-time embedded system, which consists of different processors without a common timebase, where the tasks that take place are considered as black boxes. The tasks can be, for example, signal processing algorithms. The thesis aims to analyze theoretically the origin and the magnitude of the delays and develop statis- tical methods for accurate estimations of these delays. The methods are applied to a case-study which consists of an active safety system used in Volvo Cars Corporation. The system includes camera and radar sensors and two processors which process the signals coming from the sensors and make decisions on imminent threats correspond- ingly. The traffic situations and the vehicle’s signals generated at different nodes of the system in response to these situations are logged in an Ethernet-based logging system. Logged data from drivings at public roads and test tracks are used for the analysis. Due to the lack of a common timebase, the developed methods measure the delays based initially on the processors’ timebase and then on the logger’s time- base. Finally, data coming from a reference sensor system is used to measure the same delays. The overall results of the thesis indicate that the delays depend on the complexity of the traffic environment. In addition, the measurements of delays from the different methods conclude to similar results and the outcome from the reference sensor agree with the outcome from the examined sensor system, which results in a good system’s performance.

Nyckelord: camera, distributed systems, radar, safety-critical systems, sensors, synchronization, timestamps

Publikationen registrerades 2017-04-25. Den ändrades senast 2017-04-25

CPL ID: 248961

Detta är en tjänst från Chalmers bibliotek