Distributed Opportunistic Sensing and Fusion for Traffic Congestion Detection

Abstract Military data can be difficult to acquire owing to security and operational concerns, making it challenging to undertake research in 'Anticipatory Situational Understanding for Coalitions'. To help combat this, an analogous multi-modal, multi-agency, and distributed scenario has been created to support development of tools and techniques for detecting and predicting traffic congestion in the road transport network. This scenario has been chosen as data is not generally available from a single unified source; different organizations (e.g. police, general public etc.) have sensors providing information of value. In this paper, we examine the problems of: a) identifying congestion using CCTV cameras; and b) fusing this with data from other agencies. In turn this can be used to supplement official transportation feeds. This coalition approach requires sensors to carry out assistive and supplementary tasks such as 'car counting' which this paper demonstrates how these can be done using deep learning neural networks. Utilizing distributed data sources will provide approaches that are generalizable and ad-hoc therefore more appropriate to a military context. An initial four-layer architecture and tooling are set out to enable rapid formation of human/machine hybrid teams, with a focus on opportunistic and distributed processing at the edge of the network.
  • Alistair Nottle (Airbus)
  • Dan Harborne (Cardiff)
  • Dave Braines (IBM UK)
  • Moustafa Alzantot (UCLA)
  • Santiago Quintana (Airbus)
  • Richard Tomsett (IBM UK)
  • Lance Kaplan (ARL)
  • Mani Srivastava (UCLA)
  • Supriyo Chakraborty (IBM US)
  • Alun Preece (Cardiff)
Date Sep-2017
Venue 1st Annual Fall Meeting of the DAIS ITA, 2017