Abstract

Engineers and policymakers charged with asset risk-reduction, maintenance, and replacement responsibilities need objective and actionable data to characterize their inventories and guide their decisions. However, characterization and inspection of large inventories of our built environment such as utilities and major lifeline structures remain elusive partly because of their remoteness, spatial extent, and hazardous environments. This challenge is exacerbated by prevailing cumbersome and manual data collection, interrogation, and intervention processes. Recent synergistic developments in computer vision, computing, and control have led to autonomous systems that have revolutionized many areas, including autonomous driving and the internet of things (IoT). Such systems also hold tremendous potential for asset inspections, characterization, and intervention by enabling rapid geospatial and acoustic data acquisition, data processing, and knowledge extraction on time, without relying excessively on manual processes to do so.

In this talk, the speaker will present several examples of sensing systems and the theoretical underpinnings developed by his team to address the problem of asset characterization and inspections. These include terrestrial and aquatic robots which can perceive the environment to build 3D maps of assets, e.g., nuclear and transportation. In addition to the hardware, localization, and mapping components, the speaker will also describe the probabilistic models and machine learning algorithms to extract actionable information from data using such platforms. Additionally, the speaker will present other IoT systems developed by his team, e.g., an IoT system that can continuously monitor the state of a water distribution network and a vibration-based condition monitoring system for low-speed rotating machinery. In these examples, the speaker will focus on practical challenges in developing sensing platforms while underscoring the specific infrastructure management- related problems that data from such platforms help address.