Optimizing quality-of-information in cost-sensitive sensor data fusion

Abstract

Abstract-This paper investigates maximizing quality of information subject to cost constraints in data fusion systems. We consider data fusion applications that try to estimate or predict some current or future state of a complex physical world. Examples include target tracking, path planning, and sensor node localization. Rather than optimizing generic network-level metrics such as latency or throughput, we achieve more resourceefficient sensor network operation by directly optimizing an application-level notion of quality, namely prediction error. This is done while accommodating cost constraints. Unlike prior costsensitive prediction/regression schemes, our solution considers more complex prediction problems that arise in sensor networks where phenomena behave differently under different conditions, and where both ordered and unordered prediction attributes are used. The scheme is evaluated through real sensor network applications in localization and path planning. Experimental results show that non-trivial cost savings can be achieved by our scheme compared to popular cost-insensitive schemes, and a significantly better prediction error can be achieved compared to the cost-sensitive linear regression schemes.

A Genetic Programming Approach to Cost-Sensitive Control in Wireless Sensor Networks

“…Perhaps most related to this thesis is that of Wang et al (25). They propose a method to find the optimal set of sensors to be polled, using a hybrid tree, where non-leaf nodes act as a decision tree and leaves are standard regression models using a subset of sensors.…”

A Genetic Programming Approach to Cost-Sensitive Control in Resource Constrained Sensor Systems

“…Perhaps most related to our work is that of Wang et al [27]. They propose a method to find the optimal set of sensors to be polled, using a hybrid tree, where non-leaf nodes act as a decision tree and leaves are standard regression models using a subset of sensors.…”

Maximum likelihood analysis of conflicting observations in social sensing

“…Some early applications include CenWits [Huang et al 2005], CarTel [Hull et al 2006], and BikeNet [Eisenman et al 2007]. More recent work has focused on addressing new challenges emerging in social sensing applications such as preserving privacy of participants [Ahmadi et al 2010;], improving energy efficiency of sensing devices [Nath 2012;Park et al 2011], and building general models in sparse and multidimensional social sensing space [Ahmadi et al 2011;Wang et al 2011b]. Examples include privacy-aware regression modeling, a data fusion technique that produces the same model as that computed from raw data by properly computing noninvertible aggregates of samples [Ahmadi et al 2010].…”

“…Sparse regression cube is a modeling technique that combines estimation theory and data mining techniques to enable reliable modeling at multiple degrees of abstraction of sparse social sensing data [Ahmadi et al 2011]. A further improved model to consider the data collection cost was proposed in Wang et al [2011b]. Moreover, social sensing is often organized as "sensing campaigns" where participants are recruited to contribute their personal measurements as part of a large-scale effort to collect data about a population or for some mutual interests.…”