Ignacio Solis

In-network aggregation trade-offs for data collection in wireless sensor networks

This paper explores in-network aggregation as a power-efficient mechanism for collecting data in wireless sensor networks. In particular, we focus on sensor network scenarios where a large number of nodes produce data periodically. Such communication model is typical of monitoring applications, an important application domain sensor networks target. The main idea behind in-network aggregation is that, rather than sending individual data items from sensors to sinks, multiple data items are aggregated as they are forwarded by the sensor network. Through simulations, we evaluate the performance of different in-network aggregation algorithms, including our own cascading timers, in terms of the trade-offs between energy efficiency, data accuracy and freshness. Our results show that timing, that is, how long a node waits to receive data from its children (downstream nodes in respect to the information sink) before forwarding data onto the next hop (toward the sink) plays a crucial role in the performance of aggregation algorithms for applications that generate data periodically. By carefully selecting when to aggregate and forward data, cascading timers achieves considerable energy savings while maintaining data freshness and accuracy. We also study in-network aggregations cost-efficiency using simple mathematical models. Since wireless sensor networks are prone to transmission errors and losses can have considerable impact when data aggregation is used, we also propose and evaluate a number of techniques for handling packet loss. Simulations show that, when used in conjunction with aggregation protocols, the proposed techniques can effectively mitigate the effects of random transmission losses in a power-efficient way.

Efficient continuous mapping in sensor networks using isolines

This paper introduces an energy-efficient data collection technique that takes advantage of spatial/temporal data correlation to generate maps for continuous monitoring (e.g., of environmental conditions such as temperature, humidity, etc.). In its essence, the proposed technique, isoline aggregation, works by detecting isolines which are the lines that make up a contour map. Energy efficiency through spatial aggregation is achieved by having only nodes that detect the isoline report to the sink. Simulation results show that isoline aggregation can reduce the amount of bytes transmitted by a factor of 11 compared to when no data aggregation is used and by up to 4 times when compared to an existing spatial-correlation based aggregation mechanism. At the same time, we are able to keep high data accuracy.

Isolines: energy-efficient mapping in sensor networks

This paper introduces a novel energy efficient data aggregation algorithm that targets spatially correlated data in sensor networks. Isolines aggregation works by detecting isolines which are the lines in a contour map. Energy efficiency is achieved by having only the nodes that detect the isoline report to the sink. Simulation results show that isoline aggregation can lead to significant energy savings (some scenarios reported that no aggregation can send close to 150% more bytes than isolines aggregation) with adequate data accuracy. We also compared isolines against polygon aggregation, our implementation of an approach representing existing spatially-correlated data aggregation mechanisms. Our results report that isolines exhibit higher accuracy with a slight advantage in energy efficiency.

FLIP: a flexible interconnection protocol for heterogeneous internetworking

This paper describes the Flexible Interconnection Protocol, or FLIP, whose main goal is to allow interconnection of heterogeneous devices with varying power, processing, and communication capabilities, ranging from simple sensors to more powerful computing devices such as laptops and desktops. The vision is that FLIP will be used to interconnect such devices forming clouds in the farthest branches/leaves of the Internet, while still providing connectivity with the existing IP-based Internet infrastructure. Through its flexible, customizable headers FLIP integrates just the functions required by a given application and that can be handled by the underlying device. Simple devices like sensors will benefit from incurring close to optimal overhead saving not only bandwidth, but, more importantly, energy. More sophisticated devices in the cloud can be responsible for implementing more complex functions like reliable/ordered data delivery, communication with other device clouds and with the IP infrastructure.FLIP is designed to provide a basic substrate on which to build network- and transport-level functionality. In heterogeneous environments, FLIP allows devices with varying capabilities to coexist and interoperate under the same network infrastructure. We present the basic design of FLIP and describe its implementation under Linux. We also report on FLIPs performance when providing IPv4 and IPv6 as well as transport-layer functionality a la TCP and UDP. We show FLIPs energy efficiency in different sensor network scenarios. For example, we use FLIP to implement the directed diffusion communication paradigm and obtain an improvement of 50% in energy savings over an existing directed diffusion implementation. Finally, we showcase FLIPs flexibility by demonstrating its ability to incorporate new protocol functions seamlessly. In particular, we add data aggregation functionality onto FLIP and show that it significantly increases the systems energy efficiency.

An Absration Layer for Neighborhood Discovery and Cross-Layer Metrics

Topology-awareness is an important feature of communication algorithms and protocols for wireless multi-hop networks. In its basic form, topology-awareness requires nodes to know their immediate neighbors and, if possible, the quality of the links between them. In general, neighborhood discovery is done by routing protocols; the problem with such an approach is that fundamental changes in node characteristics require revisiting the routing protocol, especially if this latter relies on cross-layer metrics. In this paper, we propose an abstraction layer that performs neighborhood discovery and link quality assessment. In this way, neighborhood discovery is decoupled from the upper layers. Our abstraction layer stands as an active building block for the implementation of routing protocols or any other IEEE 802.11 communication algorithm for wireless multi-hop networks. It provides an updated list of neighbors, along with several statistics about link quality that can be used, for instance, to compute cross-layer metrics.

The case for a flexible-header protocol in power constrained networks

This paper evaluates FLIP, a flexible header protocol for power-constrained, heterogeneous networks. We show that FLIP can improve the energy efficiency of a system considerably, prolonging the systems lifetime. For example, when employing FLIP to implement an existing sensor network communication paradigm, we obtain 50% energy savings when compared to the paradigms original implementation. Further, we use a sample data gathering application that calculates the average of a sensor networks attribute (e.g., average temperature) and show FLIPs energy efficiency when compared to static header approaches.