A framework for modeling, simulation, and automatic code generation for sensor network applications

introduction

When Wireless Sensor Networks (WSN) are used in the field, information sharing and collaborative processing are performed through wireless channel communication.

There are currently two types of functional analysis packages for WSN applications, namely platform-specific operating systems and tailor-made models, both of which have obvious defects in the application development stage. The first type is difficult to migrate applications to other platforms, such as migrating from TinyOS to MANTIS or migrating to ZigBee-compatible platforms, and a large amount of code must be rewritten and debugged. The second type has a lot of platform-dependent specific code that needs to be developed and debugged. If problems are found after the development and integration phase is over, the only solution is to debug based on LEDs, which is extremely time-consuming.

Reference Simulation Framework

The MathWorks company built a sensor network model based on Simulink as a base time simulator for dynamic systems. The model has two main components: a wireless channel model (hereinafter referred to as Communication Medium) and a Node block.

The Communication Medium block can be implemented in C and can be modified to reuse any existing channel and connection model.

The Node block has multiple fully parameterized motes, including hardware and software platform models. For example, it includes timers, LEDs, and parameterized Stateflow blocks for applications. Stateflow is a finite state machine language used to model and simulate control logic and flow charts. It can be used to implement algorithms (applications, middleware, or device drivers) running in each mote.

Simply modify the template algorithm (Stateflow library object), set the connection parameters of the nodes in the Communication Medium block, and start simulating. Collect statistics using animated state diagrams, observation instruments, and displays to perform functional analysis of the algorithm. If the analysis of the results indicates that improvements are needed, you can achieve them by improving the algorithm.

After completing the functional analysis of the algorithm, the algorithm's Stateflow is used to automatically generate code for any supported WSN platform. This is done using a custom feature of the RealTime Workshop Embedded Coder software that generates embeddable ANSI C code for Simulink and Stateflow blocks.

Figure 1 A simple simulation framework for WSN

WSN Application Operating System Examples: TinyOS and MANTIS

The programming model of TinyOS is component-based. Conceptual entities are represented by two types of components: modules and configurations. Interfaces are implemented to declare command and event signatures. Events are software abstractions, such as receiving a packet of information, completing a sensor sampling, etc. Commands are used to trigger actions, such as sensor reading operations or starting a radio to receive and transmit information. TinyOS applications are written using nesC, an extension of the C language.

MANTIS is a lightweight, multi-threaded operating system that can perform multitasking on energy-constrained distributed sensor networks. The scheduler uses a priority-based round-robin scheduler that supports thread pre-emption, allowing the operating system to switch freely between active threads without waiting. Due to this approach, the operating system responds to critical events faster than TinyOS. The MANTIS kernel and API are written in standard C.