Context-based information fusion

http://dx.doi.org/10.1016/j.inffus.2014.02.001 1566-2535/ 2014 Elsevier B.V. All rights reserved. ‘‘Context-awareness’’ might have entered the mainstream language and gained popular attention thanks to the ever increasing adoption of smart portable devices making their way into every activity of our lives. This has happened to the point that the concept of ‘‘smart’’ app (or device) has been paired if not surpassed by the concept of ‘‘context-aware’’ app (or device). The importance of exploiting contextual information (CI) for adaptively react to the surrounding environment is something that has been widely researched in the fields of distributed and mobile computing [1]. In the past few years it has become more and more apparent that fusing data from multiple sources is probably not enough. Even combining a potentially large number of sensory sources could lead to some surprises if the value being estimated, the error characteristics of the sources, and the fusion process itself, are not properly referenced to their current relevant context. For example, the state of the target is very likely to have some form of dependence on the environment surrounding the target, on the relationships and interaction with nearby entities, maybe the time of the day and weather conditions. Think for example of the case of estimating position and speed of a car driving through city traffic. The state of the car is going to be influenced by bends and turns of the road, condition of the asphalt, traffic signs, state of preceding and following vehicles, traffic rules, road works, overall traffic conditions, time of the day, weather conditions, and so on. As already understood in the mobile computing world since at least fifteen years, context is more than just location [2]. However, a clear definition of it is far from being commonly agreed upon. Mostly because of the diversity of the research fields in which the term ‘‘context’’ is picking up popularity. Very loosely speaking, context-awareness involves considering, representing, and exploiting information and knowledge that does not characterize the focal element(s) of interest but the surrounding environment or current situation. The understanding and principled exploitation of context in Information Fusion (IF) systems is still rather limited. Domain knowledge has been traditionally acquired ad hoc from experts or static archives and applied to stove-piped solutions that could hardly scale or adapt to new conditions. However, ‘‘context’’, considered as locally relevant (and possibly dynamic) domain knowledge, should play a key role at any level of a modern fusion system to gain adaptability and improved performance. Following the positive response that special sessions on context-based IF at the International conference on Information Fusion have received in the past few years, this special issue aims at providing an outlook on current research, bringing together papers covering many aspects of the inclusion of contextual elements at different levels of the fusion process. 1. Papers in this issue