WSN Management

Early deployment experiences have indicated the need for an application-cooperative management system for real-time and postmortem diagnosis of sensing applications. The Sensor Network Management System (SNMS) provides a separate networking stack for the collection and dissemination of management data, a runtime query system for the representation and retrieval of management information, and an event logging system for storing system-generated alerts. SNMS is designed to have a minimal impact on memory and bandwidth constrained TinyOS applications.

Given the requirements of a sensing application, the system addresses issues of coverage, cluster formation, and in-network aggregation by assigning roles (e.g. gateway, or aggregator) to sensor nodes. Requirements are expressed in a high-level rule specification language and they are propagated to every sensor. Rules are evaluated at the node level, by the role assignment algorithm, based on the properties of the sensor node and its neighbors.

Impala is middleware architecture for sensor application adaptability and updates. The system consists of three components, the application adapter, the application updater and an event filter. Sensors are loaded with multiple application protocols. Upon an event, e.g. a device failure, the application adapter makes a switching decision among the protocols if a property violation is detected. Application software management is based on versioning. Nodes are exchanging their software version information locally. If a node has an older version of the application, it requests the new updates from its neighbors. By employing an on-demand approach to software updates, Impala alleviates the network from unwanted traffic.

It is crucial to monitor and diagnose the health of a sensing application during its operation. The paper presents SINA, a sensor information networking architecture for fault management. The manager node is a potential bottleneck in the network, due to simultaneous incoming responses to diagnostic queries. To overcome the response implosion problem, SINA supports three operations, the sampling, self-orchestrated, and diffused computation operation. In sampling, nodes decide to respond to a query based on probabilities. The self-orchestrated operation is based on node scheduling; each responding node introduces a delay period before sending a reply. Finally, diffused computation is based on in-network aggregation, where diagnostic logic is pushed to the end nodes using mobile scripts.

The paper proposes a bandwidth management architecture to address issues of bandwidth allocation, congestion control, and gateway selection. Bandwidth allocation is based on a rule system that maps packets to a desired reception rate and a traffic class. Packets are routed based on their class priority. In contrast to end-to-end congestion control, the paper proposes a hop-by-hop flow control using synchronous negative acknowledgments (NACKs) between the receiving and the transmitting node. To prevent a gateway node from becoming a bottleneck, route selection opt for paths with high end-to-end packet success rate and low traffic load.

Topology management is important during the deployment phase of a sensing application. Various approaches assume the known position of a few nodes, e.g. some nodes may have GPS capabilities. Such approaches are usually centralized and introduce a large communication overhead. This paper proposes a distributed, GPS-free, cluster-based positioning system. Initially, a coordinate system is established locally, within a cluster, based on triangulation. The convergence of the individual systems form the global coordinate system.

Key management functions are essential for the security of sensor networks. Key management schemes can be either static or dynamic. In static key management schemes, in constrast to dynamic schemes, key distribution occurs only once, during the network initialization. The paper proposes a dynamic key management scheme based on key polynomials to prevent node capture attacks. The proposed system is based upon Exclusion Basis Systems (EBS) to secure the key re-distribution process.

The paper proposes a routing algorithm that minimizes the overall energy consumption while providing an acceptable level of performance. The proposed system assumes a cluster-based network. Each gateway node is responsible for maintaining an energy model for a subset of sensor nodes. During the configuration phase, the gateway calculates the new routes by solving a typical path-optimization problem. Routing tables are then propagated to each node. In addition to the energy-aware routing protocol, the paper also introduces a new TDMA-based MAC protocol, using breadth and depth techniques to address the slot assignment problem.

WSN Verification

In this work an energy preserving method to build/maintain the Energy Map of a WSN is proposed. In a standard approach to build/maintain an Energy Map of the WSN, each sensing-node periodically send an energy notification message to the sink-node containing its own energy level. The proposed method is aimed to reduce the number of energy-notification messages sent by the sensing-nodes to the sink, by attaching a consumption-rate to each energy-notification message, which indicates the rate at which each sensing-node is going to consume energy in future. Clearly each notification-message containing the rate information is larger than the simpler one without, however the a payoff is gained by the reduction of the refresh frequency at which such (larger) messages are sent. Each node exploits a Markovian model to derive an estimate of its consumption-rate based on observing its own behaviour. The accuracy of such an estimate is proportioinal to the duration of the observation, hence the longer the observation the more precise the estimate.

In this work Routing protocols for Tree discovery in a WSN are considered. Such protocols are aimed to determine a tree of routing paths to be used for diffusion of data from PRODUCER nodes towards CONSUMER nodes. Two approaches are distinguished: PUSH protocols as opposed to PULL protocols. With a push protocol a producer node starts the tree discovery process by sending out an exploration message, which has then to be reinforced by the consumer