Results 11  20
of
73
Robust messagepassing for statistical inference in sensor networks
 IN: PROCEEDINGS OF THE 6TH INTERNATIONAL CONFERENCE ON INFORMATION PROCESSING IN SENSOR NETWORKS IPSN’07
, 2007
"... Largescale sensor network applications require innetwork processing and data fusion to compute statistically relevant summaries of the sensed measurements. This paper studies distributed messagepassing algorithms, in which neighboring nodes in the network pass local information relevant to a glob ..."
Abstract

Cited by 12 (1 self)
 Add to MetaCart
Largescale sensor network applications require innetwork processing and data fusion to compute statistically relevant summaries of the sensed measurements. This paper studies distributed messagepassing algorithms, in which neighboring nodes in the network pass local information relevant to a global computation, for performing statistical inference. We focus on the class of reweighted belief propagation (RBP) algorithms, which includes as special cases the standard sumproduct and maxproduct algorithms for general networks with cycles, but in contrast to standard algorithms has attractive theoretical properties (uniqueness of fixed points, convergence, and robustness). Our main contribution is to design and implement a practical and modular architecture for implementing RBP algorithms in real networks. In addition, we show how intelligent scheduling of RBP messages can be used to minimize communication between motes and prolong the lifetime of the network. Our simulation and Mica2 mote deployment indicate that the proposed algorithms achieve accurate results despite realworld problems such as dying motes, dead and asymmetric links, and dropped messages. Overall, the class of RBP provides provides an ideal fit for sensor networks due to their distributed nature, requiring only local knowledge and coordination, and little requirements on other services such as reliable transmission.
Decentralised cooperative localisation for heterogeneous teams of mobile robots
 In IEEE International Conference on Robotics and Automation
, 2011
"... Abstract—This paper presents a distributed algorithm for performing joint localisation of a team of robots. The mobile robots have heterogeneous sensing capabilities, with some having high quality inertial and exteroceptive sensing, while others have only low quality sensing or none at all. By shari ..."
Abstract

Cited by 10 (1 self)
 Add to MetaCart
(Show Context)
Abstract—This paper presents a distributed algorithm for performing joint localisation of a team of robots. The mobile robots have heterogeneous sensing capabilities, with some having high quality inertial and exteroceptive sensing, while others have only low quality sensing or none at all. By sharing information, a combined estimate of all robot poses is obtained. Interrobot rangebearing measurements provide the mechanism for transferring pose information from welllocalised vehicles to those less capable. In our proposed formulation, high frequency egocentric data (e.g., odometry, IMU, GPS) is fused locally on each platform. This is the distributed part of the algorithm. Interrobot measurements, and accompanying state estimates, are communicated to a central server, which generates an optimal minimum meansquared estimate of all robot poses. This server is easily duplicated for full redundant decentralisation. Communication and computation are efficient due to the sparseness properties of the informationform Gaussian representation. A team of three indoor mobile robots equipped with lasers, odometry and inertial sensing provides experimental verification of the algorithms effectiveness in combining location information. I.
Essentia: Architecting Wireless Sensor Networks Asymmetrically
"... as one of guiding principles to architect sensor network systems.Wedemonstrateitsgenericapplicabilityandeffectivenessbyapplyingthisprincipletothreetypicalsensornetwork technologies,namely,localization(Spotlight),sensing(uSense)and communication(mNets).Thesetechnologieshaveverydissimilar features,rep ..."
Abstract

Cited by 7 (2 self)
 Add to MetaCart
(Show Context)
as one of guiding principles to architect sensor network systems.Wedemonstrateitsgenericapplicabilityandeffectivenessbyapplyingthisprincipletothreetypicalsensornetwork technologies,namely,localization(Spotlight),sensing(uSense)and communication(mNets).Thesetechnologieshaveverydissimilar features,representingawidespectrumofsystemdesignrequirements.Wehaveinvestedsignificantefforttodesign,implement andevaluateourtechniquesonTinyOS/Motetestbeds.Theresults fromseveralrunningsystemsindicatethatasymmetricfunction placementisapowerfulguidingprincipletoachieveefficiencyand highperformancesimultaneouslyinwirelesssensornetworks.At theend,weexamthesystemfeaturesthatdiscouragetheuseof asymmetricfunctionplacementandapproachestoaddressthem. I.
Efficient PeertoPeer Belief Propagation ⋆
"... Abstract. In this paper, we will present an efficient approach for distributed inference. We use belief propagation’s messagepassing algorithm on top of a DHT storing a Bayesian network. Nodes in the DHT run a variant of the spring relaxation algorithm to redistribute the Bayesian network among the ..."
Abstract

Cited by 6 (2 self)
 Add to MetaCart
(Show Context)
Abstract. In this paper, we will present an efficient approach for distributed inference. We use belief propagation’s messagepassing algorithm on top of a DHT storing a Bayesian network. Nodes in the DHT run a variant of the spring relaxation algorithm to redistribute the Bayesian network among them. Thereafter correlated data is stored close to each other reducing the message cost for inference. We simulated our approach in Matlab and show the message reduction and the achieved load balance for random, treeshaped, and scalefree Bayesian networks of different sizes. As possible application, we envision a distributed software knowledge base maintaining encountered software bugs under users ’ system configurations together with possible solutions for other users having similar problems. Users would not only be able to repair their system but also to foresee possible problems if they would install software updates or new applications. 1
Decentralised Coordination of Continuously Valued Control Parameters using the MaxSum Algorithm
, 2009
"... In this paper we address the problem of decentralised coordination for agents that must make coordinated decisions over continuously valued control parameters (as is required in many real world applications). In particular, we tackle the social welfare maximisation problem, and derive a novel contin ..."
Abstract

Cited by 6 (1 self)
 Add to MetaCart
In this paper we address the problem of decentralised coordination for agents that must make coordinated decisions over continuously valued control parameters (as is required in many real world applications). In particular, we tackle the social welfare maximisation problem, and derive a novel continuous version of the maxsum algorithm. In order to do so, we represent the utility function of the agents by multivariate piecewise linear functions, which in turn are encoded as simplexes. We then derive analytical solutions for the fundamental operations required to implement the maxsum algorithm (specifically, addition and marginal maximisation of general nary piecewise linear functions). We empirically evaluate our approach on a simulated network of wireless, energy constrained sensors that must coordinate their sense/sleep cycles in order to maximise the systemwide probability of event detection. We compare the conventional discrete maxsum algorithm with our novel continuous version, and show that the continuous approach obtains more accurate solutions (up to a 10 % increase) with a lower communication overhead (up to half of the total message size).
Stochastic Decision Making for General State Space Models
, 2009
"... Many problems in control and signal processing can be formulated as sequential decision problems for general state space models. However, except for some simple models one cannot obtain analytical solutions and has to resort to approximation. In this thesis, we have investigated problems where Sequ ..."
Abstract

Cited by 5 (3 self)
 Add to MetaCart
(Show Context)
Many problems in control and signal processing can be formulated as sequential decision problems for general state space models. However, except for some simple models one cannot obtain analytical solutions and has to resort to approximation. In this thesis, we have investigated problems where Sequential Monte Carlo (SMC) methods can be combined with a gradient based search to provide solutions to online optimisation problems. We summarise the main contributions of the thesis as follows. Chapter 4 focuses on solving the sensor scheduling problem when cast as a controlled HiddenMarkov Model. We consider the case in which the state, observation and action spaces are continuous. This general case is important as it is the natural framework for many applications. In sensor scheduling, our aim is to minimise the variance of the estimation error of the hidden state with respect to the action sequence. We present a novel SMC method that uses a stochastic gradient algorithm to find optimal actions. This is in contrast to existing works in the literature that only solve approximations to the original problem. In Chapter 5 we presented how an SMC can be used to solve a risk sensitive control problem. We adopt the use of the FeynmanKac representation of a controlled Markov chain flow and exploit the properties of the logarithmic Lyapunov exponent, which lead to a policy gradient solution for the parame
Generalizing DPOP: ActionGDL, a new complete algorithm for DCOPs
 In proceedings of AAMAS Workshop on Optimisation in MultiAgent Systems
, 2009
"... In this paper we made three main contributions (fully detailed in [5]). Firstly, we formulate a new algorithm, the socalled ActionGDL, that takes inspiration from GDL [1], extending and applying it to Distributed Constraint Optimization Problems (DCOPs). Secondly, we show the generality of Action ..."
Abstract

Cited by 4 (2 self)
 Add to MetaCart
(Show Context)
In this paper we made three main contributions (fully detailed in [5]). Firstly, we formulate a new algorithm, the socalled ActionGDL, that takes inspiration from GDL [1], extending and applying it to Distributed Constraint Optimization Problems (DCOPs). Secondly, we show the generality of ActionGDL showing how it generalizes DPOP[4], one of the low complexity, stateoftheart algorithm to solve DCOPs. Finally, we provide empirical evidence of how ActionGDL can outperform DPOP in terms of the amount of computation, communication and parallelism.
On Localization Performance in Imaging Sensor Nets
"... We propose a massively scalable “imaging ” architecture for sensor networks, in which sensor nodes act as “pixels ” that electronically reflect (and possibly modulate data on top of) a beacon transmitted by a collector node. The collector employs sophisticated radar and image processing techniques t ..."
Abstract

Cited by 4 (1 self)
 Add to MetaCart
(Show Context)
We propose a massively scalable “imaging ” architecture for sensor networks, in which sensor nodes act as “pixels ” that electronically reflect (and possibly modulate data on top of) a beacon transmitted by a collector node. The collector employs sophisticated radar and image processing techniques to localize the responding sensor nodes, and (if data modulation is present) multiuser data demodulation techniques to extract the data sent by multiple sensors. The sensors do not need to know their own locations, do not need to communicate with each other, and can be randomly deployed. In this initial exposition, we develop basic insight into the localization capabilities of this approach, ignoring sensor data modulation. This reduces to an idealized onebit, onoff keyed, communication model in which the the sensors are either “active ” or “inactive”, with the active sensors responding to the collector’s beacon without superimposing data modulation. We consider a moving collector, with the sensor reflections creating a Synthetic Aperture Radar (SAR)like geometry. However, the collector must employ significant modifications to SAR signal processing for estimation of the location of the active sensors: noncoherent techniques similar to those in noncoherent radar tomography to account for the lack of carrier synchronization between sensor and collector nodes, and decision feedback mechanisms for estimation of the locations of multiple closely spaced active sensors. Measures for localization performance are defined, and the effect of system parameters such as bandwidth, beamwidth and signaltonoiseratio on performance is investigated.
Decentralised Data Fusion: A Graphical Model Approach
"... Abstract – This paper proposes the use of graphical models to describe decentralised data fusion systems. The task of decentralised data fusion is considered as a specific instance of the general distributed inference problem in which there is a single common state of interest which is (partially) o ..."
Abstract

Cited by 3 (1 self)
 Add to MetaCart
(Show Context)
Abstract – This paper proposes the use of graphical models to describe decentralised data fusion systems. The task of decentralised data fusion is considered as a specific instance of the general distributed inference problem in which there is a single common state of interest which is (partially) observed by a number of sensor platforms. Our objective is to model and solve this problem using standard graphical model techniques. Two options for modeling the problem are considered. The model based on distributed variable cliques is found superior to a graphical model with cloned variables. The model and the messages arising through inference are compared with the wellknown Channel Filter algorithm. Our approach to inference is to apply a distributed version of the Junction Tree algorithm developed by Paskin and Guestrin. The algorithms were validated in a series of simulated tracking problems.