USACE Grant for Automating Flood Control System Inspection with Robotics and AI

Background:

Levees are an integral part of the U.S. infrastructure system that prevent flooding of numerous communities, industries, and ecosystems throughout the U.S. Currently, there are over 24,000 miles of levees recorded in the National Levee Database with significantly more levees left to inventory. Unfortunately, engineering records and instrumentation data for these levee systems is usually quite limited. As a result, inspections and assessments of levees form the primary basis for conducting risk assessments of these structures. 

Levee inspections include tasks aimed at identifying potential failure modes. Erosion and overtopping are critical failure modes for levees. Slope stability is rarely a driving failure mode. Erosion is not necessarily observable in the absence of a flooding event, during which the presence of water obscures the observability of failure indicators, complicating levee inspection processes. In addition to identifying defects and failures, inspection of levees and structures serve to create necessary information for condition and risk assessments of levee systems. The types and densities of vegetation, location of discontinuities, damage, and geometry are useful for the assessments and must be gathered via inspection, which requires significant cost and time. Inspection of culverts and other structures along levees, locations of potential critical failure modes, are more able to identify indicators of developing failure modes but are equally time consuming and costly. It is also not always known where culverts are inside of levee systems, which can be a major issue as concentrated leak erosion typically occurs along these soil-structure interfaces. Culverts from 3” to 6’ diameter are frequently inspected using robotic instruments with cameras. The videos from these culvert inspections are reviewed by human visual inspection at great time and cost. Methods are needed to gather information for assessments of levees, structures, and culverts, as well as methods to identify indicators of future failures which are rapid and affordable.

Brief Description of Anticipated Work:

The purpose of this research is to implement technologies for automating inspection methods for levees and flood control structures using robotic platforms and artificial intelligence techniques to increase accuracy, reduce time and cost, and to increase safety of performing necessary data gathering and interpretation activities. To fulfill the purpose of this research, two lines of effort will be conducted in the first year: rapid culvert inspection and rapid risk assessment data gathering for levees. Other applications will be considered for follow-on years if funding becomes available. Products to be delivered include all data sets used for the research and a final report in electronic format. All algorithms developed must be done so as to run on USACE computational resources for government personnel access and all inspection hardware and software platforms must conform to all USACE and Army cybersecurity guidelines, especially any unmanned aerial vehicles, if used. Demonstrations will be performed using data provided by the government or gathered at sites determined and coordinated by the government.