FA864924P0503

Small Business Innovation Research Program (SBIR) Phase II

Long Range, Multi-Agent GPS-Denied Autonomy, Navigation, Alt-PNT ASPN & Mod Payload Compliant Solution

The proposed effort enhances a USSOCOM funded no-emit, vision-based GPS-Denied UAV navigation solution with increased range and altitude, deploys the solution through a software download or modular payload that is compliant with Mod Payload AD1167779 and ASPN/pntOS standards, and adapts the solution to provide robust Alt-PNT to multi-agent systems. The proposed effort demonstrates the solution on a Firestorm Labs, L3Harris, Lockheed Martin UAV, or stakeholder selected UAV and delivers a hardware prototype. The solution is a vision-navigation system that uses a drone's cameras and IMUs (inertial measurement units). The solution navigates using the features in the environment, no pre-loaded terrain map needed. By combining cameras, IMU data, pixel-movement tracking, and object+feature tracking over time through Rhoman's unique system, the UAV always knows exactly where it is relative to launch. The UAV can perform any autonomous mission as if it had GPS, and when you know your location at launch, the UAV always knows it's absolute location. Deploys by Software or Hardware: The solution can deploy by software using the existing cameras and compute on the drone, or through a Mod Payload AD1167779 compliant payload: just mount to the standard payload attachment interface, perform one-time calibration manuevers post install, and the UAV is good to go. Visual Inertial Odometry + Corrector Module: The solution works by using the cameras and IMUs to enact a visual-inertial-odemetry (VIO) system to calculate the flight path of the UAV, then uses an error-correction model to cancel out and correct any drift that occurs in the VIO system. Drift Corrector Module: The drift error correction module detects objects & visual-features in the environment, tracks the relative movement of the drone versus those objects/features over time, and by measuring it's distance to multiple objects at once, over time, it determines it's actual exact movement, and corrects the VIO estimate. Object-Location-Maps, Always Return to Launch: The solution allows a UAV to always get back to launch. During flight, when the system detects and tracks objects and visual-features in the environment (bridge, group of trees, mountain ridgeline, rock formation, lake, road- intersection, edge of large farm-field, etc) as part of the drift-correction module, it also saves the description and estimated location of these objects into an object-location-map'. This map provides the UAV with visual-breadcrumbs' it can use to follow and always get back to launch. The system can also work if the UAV lands at a different location: even if it doesn't follow the visual-breadcrumbs' back to launch, it still has its Alt-PNT position estimate throughout the whole mission, and can perform missions and land in new places.

The goal of phase II is to continue the R&D efforts initiated in Phase I. Funding is based on the results achieved in Phase I and the scientific and technical merit and commercial potential of the project proposed in Phase II.

AFX235-CSO1

FX235-CSO1-2216

X23.5

Jenna Roeche