Autonomous planning and resource allocation in reconfigurable smart sensing networks
Author(s)Long, James,Ph.D.Massachusetts Institute of Technology.
Massachusetts Institute of Technology. Department of Civil and Environmental Engineering.
MetadataShow full item record
Structural health monitoring (SHM) uses sensor data to quantitatively assess the integrity and performance of infrastructure, as a basis for extending the lifespan of aging systems. Wireless sensor networks promise to enable the installation of dense arrays of battery operated sensor nodes at dramatically lower cost than traditional wired systems. However, typical vibration based health monitoring approaches sample at high frequencies, and wirelessly transmitting the resulting large volumes of data can rapidly deplete sensor node batteries. Rather than emulating the behaviour of a traditional wired SHM system, intelligent sensor nodes can analyse vibration data within the sensor network, reducing transmission volumes and conserving battery. Coordinating, configuring, and managing the resources of networks of intelligent sensor nodes, is however a significant challenge.In this research, we first propose a new computational framework for distributed in-network processing of vibration sensing data, and develop a sensor system which implements this framework. A critical advantage of this framework is its flexibility, allowing data processing logic to be remotely reconfigured almost instantaneously. We then extend this approach, developing a resource allocation algorithm which continually tracks network resources (battery life, computational power and communication bandwidth), and ensures that in-network computations utilise these resources optimally. The efficacy of this approach is then demonstrated by deploying a wireless sensor network on a steel frame tower and conducting a series of experiments investigating the performance of the proposed framework. Lastly, we consider the problem of event detection in energy-harvesting wireless sensor networks. In SHM applications, long periods of time elapse without the occurrence of any event of interest.To conserve resources, a subset of nodes can actively listen for events, while the remainder power down. Judicious planning of the sequence of active node assignments is needed to ensure that as many nodes as possible can be reached upon the detection of an event, and that the system maintains the ability to detect events in times of low energy availability. We propose and develop a novel reinforcement learning approach to this problem, and through simulation demonstrate that strategies learned by the reinforcement learning agent outperform baseline approaches. The integration of the proposed computational framework, resource allocation algorithm and collaborative event detection strategy enables fully autonomous operation of intelligent wireless SHM systems.
Thesis: Ph. D., Massachusetts Institute of Technology, Department of Civil and Environmental Engineering, 2019Cataloged from PDF version of thesis.Includes bibliographical references (pages 137-145).
DepartmentMassachusetts Institute of Technology. Department of Civil and Environmental Engineering
Massachusetts Institute of Technology
Civil and Environmental Engineering.