SeaBed Environment Developer (SeaBED)

DARPA has a need for a large number of unique simulation models that display a diversity of environment, spatial layout, and mission focus to aid in the evaluation and training of future underwater systems. The underwater mission environment is fraught with communication difficulties, novel scenarios, difficult perception tasks, and huge operational areas. Advanced autonomy could be the solution to all of these issues. A sufficiently trained and developed AI could break the reliance on human intervention, surface communication, and GPS localization and carry out complex missions across wide swaths of seafloor, including dark, opaque, or turbulent waters. Advanced autonomy requires advanced training, and advanced training requires simulation models that can mimic real environments to a sufficient level of size and detail that missions can be productively simulated. Against the larger backdrop of DARPA’s Angler program, which recognizes and emphasizes the need for the development of advanced autonomy, it is clear that DARPA requires a scalable and reliable way to convert its existing stores of underwater environmental data into large and detailed computational simulation models. Navatek proposes to develop this capability by creating an automatic programmatic pipeline that analyzes and characterizes raw sensor data, extracts the resulting point clouds, fuses them while reconciling their resolution and precision differences, eliminates noise and redundancy of the resulting large cloud, meshes them into continuous surfaces, and augments these surfaces with environmental detail from both the original sensor data and from relevant external data sources. Phase I will consist of a research effort, a feasibility assessment, and an establishment of the steps of the automatic pipeline from raw data to finished simulation environment. Navatek proposes to demonstrate the success and utility of its effort by creating a simulation example which directly showcases the capabilities of a so-created digital underwater environment. Navatek proposes to undertake this Phase I effort with a specific focus on computational feasibility and on lowering the time, processing power, and memory needed to create, store, and utilize such environments. The Phase II effort will expand on the results of Phase I and develop a complete, functional tool for the automatic creation of undersea simulation models. The successful completion of this effort will increase the availability of geographically, environmentally, seasonally, and spatially relevant simulation environments to support future underwater system mission planning, reconstruction, sensor development, and algorithm training.