Wireless sensor networking

In a few years, networked sensors and actuators will outnumber traditional electronic appliances. They will enable a plethora of new services and applications in industrial automation, asset management, environmental monitoring, medical and transportation business, and in a variety of safety and security scenarios (see Fig. 1). In the Zurich Sensor Networks & Edge Server Software projects, IBM Research is using its expertise in wireless networking, advanced middleware concepts and embedded platform design to create innovative concepts, efficient architectures and reference designs for end-to-end enterprise solutions. A reference testbed was built, which is used to verify IBM’s end-to-end concept, evaluate performance and scalability, and which serves as a demo platform for client engagements.

An overview of the ZRL wireless sensor networking testbed is shown in Fig. 2 below: it highlights the integration of wireless sensor networks with the enterprise network/intranet to bring data generated by low-end wireless sensors to an IBM WebSphere Application Server.

The sensor units contain several types of sensors for measuring temperature, light, acceleration, angular velocity, or geo-orientation, etc. A microcontroller or programmable logic device, combined with A/D converters, collects and processes the sensor signals and assembles sensor data and control frames. A serial interface connects the sensor unit to a radio module, which converts the sensor data frames to radio messages and sends them to the gateway.

All radio modules within a given range establish a wireless sensor network (WSN). The IEEE 802.15.4 / Zigbee standard is one of the most promising candidates for designing WSNs, which need to be self-organized and self-healing, i.e. nodes automatically establish and maintain connectivity among themselves. Mesh-networking protocols provide new capabilities where each node operates not only as a direct source or sink, but also as a message forwarder for other nodes that do not have direct connectivity with their communication peers. Providing reliable wireless connectivity, stability and scalability, while at the same time coping with the limitations imposed by low-cost, battery-powered sensor nodes, presents a multitude of challenging research problems.

The sensor gateway and middleware provide the bridge between the wireless sensor world and the enterprise computing environment. Based on an advanced version of the IBM WebSphere Micro Environment running a J9 Java virtual machine, the gateway extracts sensor frames received from the WSN and publishes the sensor data to a message broker by means of a telemetry transport protocol (MQtt) and a client.

The publish/subscribe messaging model allows the data collected by the sensors to be available to enterprise applications in the same way as any other enterprise information. Sensor applications subsequently receive relevant sensor information for further processing, event filtering or alert generation. In addition, the middleware provides a platform for deployment and maintenance of services and applications on networked devices.

IBM’s sensor network testbed uses the Service Management Framework (SMF), IBM’s implementation of the Open Services Gateway Initiative (OSGi) specification, to remotely start, stop and manage message filtering applications and software packages running on the gateway without interrupting the operation of the device. To support a very large number of sensors and/or high message volumes, the architecture must be intrinsically scalable and capable of adapting dynamically to the load. Towards this goal, IBM Research has introduced a novel concept that allows distribution of the message traffic across several message channels. This is achieved in IBM’s architecture by adding a channel controller to each message channel, which dynamically tracks the state of the underlying message channel (e.g. load, topics published, topics subscribed) and reports the state by means of a meta-channel to other channel controllers and system entities. Other channel controllers listen and react, so that message traffic flows over less congested message channels. The meta-channel concept leads to a self-adjusting system, which achieves load balancing, automatic configuration and resiliency, and makes the overall architecture fully scalable. A more detailed discussion can be found in the paper “Edge Server Software Architecture for Sensor Applications.”

Many sensor and actuators are low-power devices that require only intermittent connectivity and can operate on batteries for weeks or months. Current publish/subscribe middleware has not been designed for such scenarios. Therefore, IBM Research is working on novel light-weight messaging protocols that do not rely on full TCP/IP connectivity and are capable of operating directly over low-power wireless protocols. This will allow messaging oriented middleware to be brought down to very low-end sensors or actuators enabling a true end-to-end solution.

To summarize, IBM’s research focuses on wireless technologies and protocols, hardware/software platforms for smart sensor units and gateways, and messaging middleware. Researchers carry out performance evaluation with short-range wireless technologies that are highly relevant for sensor networking, such as IEEE 802.15.4 / ZigBee, Ultra Wideband, Bluetooth, and IEEE 802.11. IBM Research also explores the merits of wireless mesh networking to extend the range and enhance reliability. In addition, the research team addresses the growing need for location-sensing functionality in wireless and sensor networks.

Project pages: Sensor Networks & Edge Server Software

Rate this article

Rate from		low...		to	...high
Usefulness: