Learning Data Delivery Paths in QoI-Aware Information-Centric Sensor Networks


Singh G. T., Al-Turjman F. M.

IEEE INTERNET OF THINGS JOURNAL, cilt.3, sa.4, ss.572-580, 2016 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 3 Sayı: 4
  • Basım Tarihi: 2016
  • Doi Numarası: 10.1109/jiot.2015.2504487
  • Dergi Adı: IEEE INTERNET OF THINGS JOURNAL
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Sayfa Sayıları: ss.572-580
  • Anahtar Kelimeler: Cognition, data delivery, information-centric, large scale, learning, quality-of-information (QoI), sensor networks, CONNECTIVITY, FEDERATION, FRAMEWORK
  • Orta Doğu Teknik Üniversitesi Adresli: Evet

Özet

In this paper, we envision future sensor networks to be operating as information-gathering networks in large-scale Internet-of-Things applications such as smart cities, which serve multiple users with diverse quality-of-information (QoI) requirements on the data delivered by the network. To learn data delivery paths that dynamically adapt to changing user requirements in this information-centric sensor network (ICSN) environment, we make use of cognitive nodes that implement both learning and reasoning in the network. In this paper, we focus on the learning strategies and propose two techniques, namely learning data delivery A* (LDDA*) and cumulative-heuristic accelerated learning (CHAL) that use heuristics to improve the success rate of data delivered to the sink in the cognitive ICSN. While LDDA* updates a single heuristic function to choose paths that can deliver data with good QoI to the sink, CHAL accumulates heuristic values from multiple observations from the environment to choose data delivery paths that are more resource aware and considerate toward the energy consumption of the network. Extensive simulations have shown improvement of about 40% in the average rate of successful data delivery to the sink with the use of heuristic learning, when compared with a network that did not implement any learning.