Intelligent architecture through a supervised learning approach in wireless multimedia sensor networks

  • Authors:

  • Honggang Wang;Shaoen Wu

  • Affiliations:

  • University of Massachusetts, Dartmouth;University of Southern Mississippi

  • Venue:
  • SpringSim '10 Proceedings of the 2010 Spring Simulation Multiconference
  • Year:
  • 2010

Quantified Score

Hi-index 0.01

Visualization

Abstract

Future multimedia applications in wireless multimedia sensor networks impose stringent and diverse Quality of Service (QoS) requirements, which demands a cross-layer architecture. However, the cross-layer optimization for multimedia delivery is not easy to be implemented in resource limited Wireless Multimedia Networks, as QoS requirements may be qualitative, or even fuzzy. In practice, QoS requirements can be defined as a set of constraints indicated by parameters including lifetime, latency, accuracy, energy efficiency, reliability, robustness and scalability. There are interactions among these parameters across layers. Previous cross-layer approaches usually search exhaustively for all possible strategies and use all assumed parameters without considering importance. The computational burden makes them difficult to apply in real practice. Thus, it is critical to identify these interactions and determine a single optimized cross-layer control strategy, where the overall system performance is not independently determined by all parameters at each individual layer, but is determined by significant parameters and their interactions of equivalent layers. In this paper, a supervised learning approach is incorporated into cross-layer architecture to reduce the number of optimized cross-layer parameters based on their significance. Our study shows that the proposed intelligent architecture can significantly reduce the associated complexity (e.g. control, information exchange and computation overheads).