lab personnel
private area
project description
proof of concept
work plan
broader impacts

  The challenge of wirelessly networking a large number of sensors stems from two facts. First, the radio communication medium is shared by all, while each sensor may only be able to communicate directly with a few of its neighbors (imagine trying to reliably send a voice message to a friend at the other side of a crowded sports arena). Secondly, the radio medium is inherently unreliable due to noise, time- and frequency selectivity, and interference. Hence the key network engineering problems are: (i) Can the networking algorithms scale up to enormous numbers of nodes? (ii) Are those algorithms robust to the unpredictable and hostile medium? (iii) Do the algorithms provide sufficient performance and energy efficiency for the application?

Driven by the environmental sensing application, our network has two operational phases: self-assembly and data collection. Self-assembly---the automatic, unattended setup of communication in the network---occurs after deployment of the sensors, and consists of route discovery and proactive coordination for access to the shared radio spectrum. Our route discovery algorithm is based that fact that most information flows are from the sensors to the gateway node for storage or forwarding over a long-haul link. Starting from the gateway, the network discovers logical concentric rings of sensor nodes indexed by the number of radio hops from the gateway. As part of the process the nodes establish cliques with neighboring nodes in adjacent rings that most efficiently (in terms of time and energy consumption) relay information toward the gateway.

Since the communication medium (or channel) is shared, a medium access control (MAC) protocol is necessary to carve the medium into chunks and efficiently allocate those chunks to radio transmissions during the data collection phase. Many MAC schemes rely on contention for the channel. While simplistic, contention-based protocols waste energy due to the collision of transmissions from neighboring nodes. Our MAC algorithm (Flikkema and West 2003) employs local proactive coordination so that a node is awake---and consuming energy---for communication only when it has a meeting scheduled with one of its cliques, significantly improving energy efficiency over contention-based schemes (see Ye et al. 2002 and the references therein). It also uses a form of time-frequency randomization of transmissions enabled by our tunable-frequency radios. Randomized MAC algorithms for wireless sensor networks are receiving much attention in the research community (Mergen and Tong 2002, Rozosky and Kumar 2001, Sohrabi and Pottie 1999) due to their good scaling properties. Our approach is unique in that it is integrated with the clique memberships established in the route discovery phase. During self-assembly, clique members share information so they can predict exactly when a meeting will occur in the data collection phase. However, different cliques use different randomized schedules so that they can communicate with little risk of interfering with each other. This approach obviates expensive coordination across the entire network, and it makes possible extremely large populations of sensors in a single network. It is also robust to electromagnetic interference from other radio-based systems as well as to dynamic variation---fading---of the wireless channel.
Northern Arizona University EnGGen Homepage