Quality of Service of Networked Embedded Systems

Project Description

Research activities in the area of “networked embedded systems in dynamic environments” aim at an investigation of the performance and real-time characteristics of such distributed systems. For this purpose, several system components with their internal behaviour (like operating system, application program, runtime system) and the communication units are examined more closely. The test scenario was specified as a variant of the RoboCup Small Size League (F180). The system consists of individual soccer robots, a global vision system for the playfield, as well as a computer system which acts as a communication base using WLAN and Ethernet. For the investigation, an adequate measurement infrastructure is necessary, which can monitor response time and utilization of the system. It is required to adjust the already existing measurement infrastructure of a web cluster at the group for embedded systems. Simulation models (evaluated by means of tools like AnyLogic, OPNET, ns-2) and analytic queueing models (using e.g., WinPEPSY) can then be built based on the measurements. Based on this measurement environment and the developed models, it is possible to validate and calibrate close-to-reality models, which allow to assess and interpret such systems. This work will answer the question “which components must the model contain to describe the system adequately”.
The work of the previous years can be summarized in four parts. In the first part, the self-built-robot has been redesigned including a microcontroller-circuit, which communicates via the RS-232-Interface of the PC/104-board. The microcontroller enables the robots of each team to act very precisely in the dynamic environment of a soccer game, for instance.
In the second part, the operating system of the PC/104-board has been patched with RTAI to analyze the performance of the real-time-application-interface for Linux. To evaluate the system response in real time, signal edges have been generated and the latency was monitored using a very highly-precise time-measuring instrument. Monitoring systems delivered a variation of the system response time when calling a single-shot-task and periodic-tasks and if the system load increases i.e. by executing classic tasks like gcc and pingflood.
In the third part, an existing event-tracing-tool for the Linux-Kernel was modified for tracing real-time events by using the RTAI-Patch for Linux. Now it is possible to analyze the latency of the response time on the system software level by tracking the real-time system calls.
In the fourth part, the software of the existing soccer-playing robot has been adapted, so this player now can act more autonomously by using its on-board-sensors (proximity-sensors and cmos camera). The robot is not provided with any kind of information from outside the system.

Project Period

2002-06-01 – 2011-12-31

Project Members

Related Publications