A real-time kernel for wireless sensor networks employed in rescue scenarios
In rescue scenarios, real-time requirements are one key issue when using wireless sensor networks (WSNs) for tracking and monitoring of rescue forces. If a node detects an alarm condition, the alarm message must be delivered to the base station in time. To guarantee the timely delivery the whole system architecture has to fulfill real-time requirements.
However, there are only a few real-time architectures which can be used for sensor networks. These architectures are too generic to comply with the other requirements of a WSN architecture, like RAM usage and energy awareness.
In this paper, we present FireKernel, a real-time micro kernel designed for WSN operating systems with a special focus on hard real-time requirements and strict energy management. It offers a preemptive real-time scheduler, mutexes and synchronous message passing for interprocess communication (IPC).
The kernel can easily be adopted to existing WSN firmware because it is transparent to the programmer and has a very simple programming model. Due to the introduction of tickless timers which are not using any periodical timer interrupt, the whole system can stay in its deepest low-power mode as long as no calculation is being done.
The scheduler is priority based and uses no periodical timers. As part of the kernel we have incorporated a tickless timer system. We have implemented and tested the kernel on MSP430 and ARM7 TDMI based sensor nodes.
We discuss the kernel’s architecture, implementation and resource usage. Further, we present benchmark results for interrupt latency and energy consumption of the timer architecture.
To read this external content in full, download the complete paper from the author archive online.