Performance analyses of embedded real-time operating systems using high-precision counters

To evaluate the performance of embedded real-time operating systems, we have built a scenario of soccer playing robots according to the F180 small-size league of the Robocup with a common embedded industrial PC/104 system. Thereafter we patched a common Linux kernel with the real-time application interface RTAI and installed it on a compact flash card for using it on the robots’. We connected IR distance sensors to the onboard embedded system and performed a response time analysis of the operating system, and implemented an interrupt service routine for the board’s parallel port to generate a system response for externally caused hardware interrupts, e.g. from sensors. For a faster data collection we triggered interrupts by edges using a function signal generator. To monitor the embedded systems’ response time we built a monitoring system by using a high-precision histogram scaler and counter. The data for the response time has been monitored in various system loads and been analysed statistically. Our system allows an easy and low-cost way for performance analyses of embedded real-time operating systems.