Yao Win Hong

Opportunistic large arrays: cooperative transmission in wireless multihop ad hoc networks to reach far distances

The technique we propose in this paper allows efficient flooding of a wireless network with information from a source, which we refer to as the leader. At the same time, it permits us to transmit reliably to far destinations that the individual nodes are not able to reach without consuming rapidly their own battery resources, even when using multihop links (the reach-back problem). The synchronization constraints are extremely loose and can be fulfilled in a distributed manner. The key idea is to have the nodes simply echo the leaders transmission operating as active scatterers while using adaptive receivers that acquire the equivalent network signatures corresponding to the echoed symbols. The active nodes in the network operate either as regenerative or nonregenerative relays. The intuition is that each of the waveforms will be enhanced by the accumulation of power due to the aggregate transmission of all the nodes while, if kept properly under control, the random errors or the receiver noise that propagate together with the useful signals will cause limited deterioration in the performance. The avalanche of signals triggered by the network leaders form the so-called opportunistic large array (OLA). The main advantages of the OLA are its great flexibility and scalability.

A scalable synchronization protocol for large scale sensor networks and its applications

Synchronization is considered a particularly difficult task in wireless sensor networks due to its decentralized structure. Interestingly, synchrony has often been observed in networks of biological agents (e.g., synchronously flashing fireflies, or spiking of neurons). In this paper, we propose a bio-inspired network synchronization protocol for large scale sensor networks that emulates the simple strategies adopted by the biological agents. The strategy synchronizes pulsing devices that are led to emit their pulses periodically and simultaneously. The convergence to synchrony of our strategy follows from the theory of Mirollo and Strogatz, 1990, while the scalability is evident from the many examples existing in the natural world. When the nodes are within a single broadcast range, our key observation is that the dependence of the synchronization time on the number of nodes N is subject to a phase transition: for values of N beyond a specific threshold, the synchronization is nearly immediate; while for smaller N, the synchronization time decreases smoothly with respect to N. Interestingly, a tradeoff is observed between the total energy consumption and the time necessary to reach synchrony. We obtain an optimum operating point at the local minimum of the energy consumption curve that is associated to the phase transition phenomenon mentioned before. The proposed synchronization protocol is directly applied to the cooperative reach-back communications problem. The main advantages of the proposed method are its scalability and low complexity.

Time synchronization and reach-back communications with pulse-coupled oscillators for UWB wireless ad hoc networks

Time synchronization has been an extremely difficult issue for wireless ad hoc networks due to its decentralized nature. Interestingly, synchrony have often been observed in swarms of biological systems such as that of synchronous flashing fireflies or spiking of neurons. In this paper, we utilize the narrow pulse characteristics of UWB systems to emulate the pulse-coupled integrate-and-fire (IF) model embedded in biological swarms in order to achieve distributed synchronization. The method is based on a simple transmission strategy where nodes integrate the coupling caused by the signal pulses received from other nodes, and fire a pulse after reaching a designated threshold. With time synchronization, many cooperative strategies can be applied to the network of distributed nodes. In particular, we show that synchronization can lead to coherent superposition of the signal pulses and it would allow to utilize the network as a distributed antenna array capable of reaching far receivers, solving the so called reach-back problem.

On multiple access for distributed dependent sources: a content-based group testing approach

In this paper we consider the multiple access problem with distributed dependent sources. We derive the optimal designs for the case of N correlated binary sources whose data is modelled as a two-state Markov chain. The solution can be classified as a group testing technique where data values at the sensors are determined through the successive refinements of the tests over smaller groups. The tests form, progressively, an accurate map of the sensor data at the central receiver. We derive the conditions on the parameters of the data model for which the group testing approach is superior to time sharing. In contrast to standard multiple access techniques, this is the first method proposed for data retrieval from distributed dependent sources which is content-based rather than user-based.

Distributed change detection in large scale sensor networks through the synchronization of pulse-coupled oscillators

This paper proposes the use of a distributed synchronization mechanism, which locks in phase the pulse-coupled oscillators, to rapidly alert the nodes in a sensor network of a change detected by a group of the sensors. By encoding into an abrupt variation of the phase their positive detection of a change, the nodes force all other nodes to reach a new synchronization equilibrium. Therefore, the information about the change is implicitly encoded in the phase transitions. While the local detection problem at each sensor can be addressed using the standard change detection algorithms, the interesting aspect of this work is the unconventional way through which the nodes broadcast their information to each other and fuse their decisions. The main advantages of the proposed method is the scalability and low complexity of the fusion algorithm.

A simple method to reach detection consensus in massively distributed sensor networks

A bio-inspired distributed detection protocol is proposed where the synchronization of pulsing devices is used to reach a consensus in the decision among sensor nodes. Each node encodes the local decision into their pulsing time and reaches an agreement when the pulsing instants converge to a common value. The protocol has low complexity and scales favorably with the size of the network. In fact, we prove that the probability of detection asymptotically goes to 1 as the number of sensors goes to infinity.

A communication architecture for reaching consensus in decision for a large network

One of the most challenging aspects in applying decentralized detection in sensor networks is the efficient exchange of small messages required for data fusion. In this work, we propose a novel communication architecture for a canonical decentralized detection problem where the sensor nodes exchange continuously their local decisions until consensus is reached among all nodes. Our methodology capitalizes on the observation that the information embedded in the exchanged messages decreases to zero as the decisions gradually converge. By using a data-driven multiple access scheme, we show that the number of channel accesses required for each round of message exchange decreases, following the same trend as the aggregate entropy of the sensor decisions. The main contribution is to show that data-driven multiple access strategies can overcome the backlog of communications that many distributed computing algorithms generate in a wireless network setting

Group testing for sensor networks: the value of asking the right questions

Sharing of information is crucial in sensor networks to allow collective processing of observations made by distributed sensors. Our goal is to efficiently retrieve the distributed sensor measurements at a central processor or to share the information among all sensors. To achieve this goal with a minimum number of channel accesses, it naturally involves the compression of the distributed source data and an optimal scheduling of the transmissions to reduce the number of redundant transmissions. In this paper, we utilize a content-based group testing approach to derive a joint source coding and multiple access scheduling method without the initial knowledge of the statistics of the sensor field. The group testing multiple access (GTMA) scheme proposed in this paper is obtained by choosing groups through a tree splitting algorithm that adapts the branching of the tree according to the progressively estimated statistics of the sensor field. We show that this method overcomes the difficulty of applying the algebraic distributed source coding schemes to a large number of sensors and for arbitrary or unknown statistics of the sensor field.

Generalized group testing for retrieving distributed information

The goal of group testing is to efficiently classify the state of a set of distributed agents through a sequence of tests by imposing each test simultaneously upon groups of agents. In this work, we describe the concept of group testing in a generalized framework and propose to apply this concept to solve the scheduling and multiple access problem in a large scale wireless sensor network. Since the standard approach is to dedicate a single channel to each sensor, we discuss the efficiency of group testing by comparing it to the case where each sensor is tested individually. Through the sequence of tests, the group testing strategy successively refines the observation space of the set of sensors and eventually identifies the status of each sensor when the space is refined to only one element. We show that the successive refinement property of group testing (similar to that of arithmetic coding) plays an important role in its performance. Based on this concept, we provide insight into choosing optimal group testing strategies for general applications.

Cooperative transmission in wireless multi-hop ad hoc networks using opportunistic large arrays (OLA)

We propose a network flooding algorithm for wireless ad hoc networks based on a pure physical layer design. The idea is to let the nodes in the network act as asynchronous relays to a specific leader, generating an avalanche of signals which we call opportunistic large array (OLA). The accumulation of power in the medium coming from the interfering signals constructs a signal space for each receiver node that can be acquired through training or blindly. The acquisition of the signal space allows the leader to flood the network without higher layer intervention. Compared to other approaches that are based on the network layer and use point-to-point links to flood the network, we show that the proposed flooding algorithm achieves higher efficiency in the end-to-end delay and has a higher delivery ratio due to the accumulation of energy. In fact, a major contribution of this paper is mapping the multiple-access network problem into a single user problem and then using signal processing techniques to solve it instead of networking algorithms.