Person:

Mainland, Geoff

This paper presents market-based macroprogramming (MBM), a new paradigm for achieving globally efficient behavior in sensor networks. Rather than programming the individual, low-level behaviors of sensor nodes, MBM defines a virtual market where nodes sell "actions" (such as taking a sensor reading or aggregating data) in response to global price information. Nodes take actions to maximize their own utility, subject to energy budget constraints. The behavior of the network is determined by adjusting the price vectors for each action, rather than by directly specifying local node actions, resulting in a globally efficient allocation of network resources. We present the market-based macro-programming paradigm, as well as several experiments demonstrating its value for a sensor network vehicle tracking application.

This paper addresses the problem of resource allocation in sensor networks. We are concerned with how to allocate limited energy, radio bandwidth, and other resources to maximize the value of each node's contribution to the network. Sensor networks present a novel resource allocation challenge: given extremely limited resources, varying node capabilities, and changing network conditions, how can one achieve efficient global behavior? Currently, this is accomplished by carefully tuning the behavior of the low-level sensor program to accomplish some global task, such as distributed event detection or in-network data aggregation. This manual tuning is difficult, error-prone, and typically does not consider network dynamics such as energy depletion caused by bursty communication patterns.

We present Self-Organizing Resource Allocation (SORA), a new approach for achieving efficient resource allocation in sensor networks. Rather than manually tuning sensor resource usage, SORA defines a virtual market in which nodes sell goods (such as sensor readings or data aggregates) in response to prices that are established by the programmer. Nodes take actions to maximize their profit, subject to energy budget constraints. Nodes individually adapt their operation over time in response to feedback from payments, using reinforcement learning. The behavior of the network is determined by the price for each good, rather than by directly specifying local node programs.

SORA provides a useful set of primitives for controlling the aggregate behavior of sensor networks despite variance of individual nodes. We present the SORA paradigm and a sensor network vehicle tracking application based on this design, as well as an extensive evaluation demonstrating that SORA realizes an efficient allocation of network resources that adapts to changing network conditions.