Self-Similar Algorithms for Dynamic Distributed Systems

2010 - Milcom 2010 Military Communications Conference

Liu

2010

Self Cite

Abstract-We study systems of static and mobile sensors in which the participating nodes are disconnected. At any given time, only selected groups of nodes are able to interact. In this paper, we restrict attention to pairwise interactions of static nodes placed on a regular grid. Nodes must agree on a schedule that specifies when and in what order these interactions take place. Our focus is the impact this interaction schedule has on systemwide properties, such as information propagation and resiliency. We show how interaction schedules with varying properties can be built from simple patterns and we discuss the choice of suitable patterns on sample illustrative scenarios. I. CONTEXT AND PROBLEM DEFINITION A. Energy constraints and intermittent connectivityThe lifespan of a sensor network is typically determined by the amount of available energy in the nodes and the rate with which the energy is consumed. Technological constraints and cost make it difficult to increase the amount of energy in the nodes. Communication is one of the most significant energy consuming activities of a sensor node. A common method to extend the lifespan of a sensor network is to have the sensors spent a significant percentage of time in low energy consumption (sleep) mode [1]. Depending on the "depth" of the sleep mode, a node might be capable of sensing but it is not capable of sending or receiving. When a node is fully awake, it can send and transmit but consumes a significant amount of power which drastically reduces its lifespan.The overall objective is to satisfy the requirements of a network's mission while minimizing the amount of energy consumed. To this end, we aim to reduce the energy cost of communication by reducing the time spent in the fully awake mode and by limiting the overhead and interference. More specifically, we eliminate the need for media access control (MAC) at runtime by orchestrating the exchanges between agents offline before deployment in a way that only two agents take part in any one exchange. Furthermore, these exchanges are scheduled (in time and possibly communication channel) so that no other node within the interference range can transmit at the same time. In such a scheme, nodes can transmit without a prior channel arbitration that is normally required to address the hidden terminal problem. This is especially important in networks with high propagation latency, such as B. A network model based on scheduled meetingsThe nature of the networks described above leads to a model in which nodes interact in local groups and do not attempt to reach specific nodes outside their group through long distance mechanisms like multi-hop routing. There are other scenarios in which this form of communication-group based and local only-is appropriate, including cases where nodes are mobile. In this paper, we focus on pairwise meetings of static sensor nodes laid out on a regular grid, but the model we propose is more general and can be applied to non regular topologies, possibly with meetings of more than two nodes...

Section: Discussionmentioning

confidence: 99%

Interaction patterns for resilient intermittently-connected static sensor networks

Charpentier

2010 - Milcom 2010 Military Communications Conference

Liu

2010

Self Cite

“…More involved functions and/or states can have a substantial impact on the design and optimization of tours. We have started to explore classes of functions that will give network designers more freedom in the way they can arrange tours [17] and we are also looking at tradeoffs between the size of agents' states and the amount of interaction needed to solve a mission with the desired performance. We have also completed preliminary work on techniques that can be used to design tour-implementing trajectories in such a way that the cost of high meeting rates is minimized.…”

Section: Current and Future Workmentioning

confidence: 99%

A Mechanism to Structure Mission-Aware Interaction in Mobile Sensor Networks

Charpentier

Distributed Computing and Networking

Bhatia

2008

Self Cite

Abstract. One of the main appeals of mobile sensors is the variety of environments in which they can operate as an autonomous network. Different environments, however, present different challenges, especially in terms of inter-sensor communication. In sparse environments, it may not be possible to maintain full connectivity at all times, setting off the need for agents to communicate opportunistically when they are close to each other. This, in turn, suggests that communication needs be taken into account in the design of agents' trajectories. In this paper, we introduce a notion of tour and meeting point as an abstraction of trajectories designed for both sensing and communication, and we study the tradeoffs involved between motion to sense and sample and motion to communicate and interact.

“…Implicit in the paper by Chandy et al [8] are the following questions: How can we derive distributed algorithms that compute correctly in dynamic environments? Which functions are amenable to distributed computation in dynamic environments?…”

Section: Self-similar Algorithmsmentioning

confidence: 99%

“…In this paper, we focus on two recent papers that deal with distributed computation in dynamic environments-the first by Chandy et al [8] and the second by Angluin et al [9]-and attempt to characterize the relationship between their work. Both papers are motivated by the need to understand computation in distributed systems that exist in highly dynamic environments similar to those in which mobile sensor networks are deployed.…”

Section: Introductionmentioning

confidence: 99%

Self-similar Functions and Population Protocols: A Characterization and a Comparison

Bhatia

Distributed Computing and Networking

2008

Abstract. Chandy et al. proposed the methodology of "self-similar algorithms" for distributed computation in dynamic environments. We further characterize the class of functions computable by such algorithms by showing that self-similarity induces an additive relationship among the level-sets of such functions. Angluin et al. introduced the population protocol model for computation in mobile sensor networks and characterized the class of predicates computable in a standard population. We define and characterize the class of self-similar predicates and show when they are computable by a population protocol.