Towards Real-Time Information Processing of Sensor Network Data using Computationally Efficient Multi-output Gaussian Processes M. A. Osborne and S. J. Roberts Department of Engineering Science University of Oxford Oxford, OX1 3PJ, UK. {mosb,sjrob}@robots.ox.ac.uk A. Rogers, S. D. Ramchurn and N. R. Jennings School of Electronics and Computer Science University of Southampton Southampton, SO17 1BJ, UK. {acr,sdr,nrj}@ecs.soton.ac.uk Abstract In this paper, we describe a novel, computationally effi- cient algorithm that facilitates the autonomous acquisition of readings from sensor networks (deciding when and which sensor to acquire readings from at any time), and which can, with minimal domain knowledge, perform a range of information processing tasks including modelling the accu- racy of the sensor readings, predicting the value of missing sensor readings, and predicting how the monitored environ- mental variables will evolve into the future. Our motivating scenario is the need to provide situational awareness sup- port to first responders at the scene of a large scale incident, and to this end, we describe a novel iterative formulation of a multi-output Gaussian process that can build and ex- ploit a probabilistic model of the environmental variables being measured (including the correlations and delays that exist between them). We validate our approach using data collected from a network of weather sensors located on the south coast of England. 1 Introduction Sensor networks have recently generated a great deal of re- search interest within the computer and physical sciences, and their use for the scientific monitoring of remote and hostile environments is increasingly common-place. While early sensor networks were a simple evolution of existing automated data loggers, that collected data for later off-line scientific analysis, more recent sensor networks typically make current data available through the internet, and thus, are increasingly being used for the real-time monitoring of environmental events such as floods or storm events (see [7] for a review of such environmental sensor networks). Such real-time access to sensor data is also a feature of pervasive sensor systems in which sensors owned by mul- tiple stakeholders (e.g. private individuals, building own- ers, and local authorities) are ubiquitously deployed within urban environments and make their information available to multiple users directly through standard web interfaces (see the CitySense project of Harvard University [13] and Microsoft’s SenseWeb project [1]). Such networks have many applications, including traffic or pollution monitor- ing, and within the ALADDIN project (http://www. aladdinproject.org), we are seeking to use such net- works to provide situational awareness support to first re- sponders at the scene of a large scale incident. We envisage providing these first responders with a mobile computer or personal digital assistant (PDA) that is capable of collect- ing information from local sensors, compiling a coherent world view, and then assisting in decision making. An ex- ample application would be to provide fire fighters with lo- cal weather information, and to predict future wind changes through observations of nearby sensors. Other applications include tracking the movement of dangerous gas, chemical or smoke plumes, and monitoring the structural integrity of buildings after an earthquake. Using real-time sensor data in this manner presents many novel challenges; not least the need for self-describing data formats, and standard protocols such that sensors can ad- vertise their existence and capabilities to potential users. However, more significantly for us, many of the information processing tasks that would previously have been performed off-line by the owner or single user of an environmental sen- sor network (such as detecting faulty sensors, fusing noisy measurements from several sensors, and deciding how fre- quently readings should be taken), must now be performed in real-time on the mobile computers and PDAs carried by the multiple different users of the system (who may have different goals and may be using sensor readings for very different tasks). Furthermore, to support decision making, it may also be necessary to use the trends and correlations observed in previous data to predict the value of environ- mental parameters into the future, or to predict the reading