ZebraNet and Beyond: Collaboration in Sparse Mobile Networks

Опубликовано 6 сентября 2016, 17:09

With the proliferation of sensor networks, the ever expanding variety of applications has driven researchers to focus not only on fixed networks, but also on mobile networks. For many reasons, both technical and logistical, such networks will often be very sparse for all or part of their operation, necessitating their need to function as disruption-tolerant networks (DTNs). In this talk, I will discuss collaboration techniques for both data gathering and parameter estimation in sparse DTNs. I will first introduce the ZebraNet system, a sparse mobile network we designed and deployed for animal tracking in areas with scarce infrastructure, and its collaborative data gathering technique for efficiently collecting animal position data. I will then devote the rest of the talk to my dissertation work that extends the collaborative technique to dynamic information sharing. I will in particular focus on LOCALE, a collaborative localization technique for low-density networks without per-node GPS. Due to the very sparse nature of DTNs, instant information sharing is impossible. The key novelty of LOCALE is that nodes not only collaborate with occasional neighbors, but also actively predict their own position through inexpensive movement tracking during periods of disconnection. Thus, LOCALE enables ΓÇ£real-timeΓÇ¥ collaborative location estimation even when networks are very sparse. I will present simulation and implementation results that show its accuracy, energy savings, and infrastructure savings compared to other prior techniques. For example, LOCALE produces accurate location estimations with 150X power savings over per-node GPS.