Fusion Flight
Advanced AI and Autonomy Sensor Fusion Project
Fusion Flight is a high-impact research project at the University of Washington's ARC Lab. We are building the next generation of autonomous aerospace systems by tackling the hard problems at the intersection of artificial intelligence, robotics, and control theory.
01
Overview
Many autonomous platforms treat positioning signals as passive inputs that must remain clean and well behaved. Fusion Flight takes a fundamentally different approach. We treat positioning waveforms and their distortions as an active sensing modality. Instead of viewing noise, multipath, and interference as nuisances, we extract the embedded structure within these effects to infer environmental conditions, signal integrity, and navigation reliability. By using noise itself as a measurement channel, Fusion Flight enables autonomy that remains reliable, precise, and predictable even when GPS is degraded, sensors are partially corrupted, or the operating environment is adversarial.
02
Problem
Autonomous aerospace systems depend on GPS for position, velocity, and timing, yet GPS signals in real environments are shaped by complex multipath interactions rather than clean line of sight propagation. Reflections from buildings, terrain, vegetation, and the vehicle’s own structure create overlapping delayed replicas of the signal that distort the correlation function in ways that are nonlinear, time varying, and difficult to model. These distortions introduce biases that standard receivers cannot reliably detect or characterize. Once introduced, these errors propagate through the navigation stack, corrupting state estimates and reducing the stability and predictability of the vehicle’s guidance and control behavior. The problem is particularly acute in dense urban areas, mountainous regions, low altitude airspace, and any environment with strong reflective or obstructive features.
The scope of this problem is broad because multipath and signal distortion are now the norm rather than the exception for UAVs, autonomous aircraft, and distributed aerospace platforms. As operations expand into environments with complex RF structure or contested signal conditions, the frequency and severity of GPS degradation will increase. If not addressed, the impact is substantial: reduced navigation integrity, increased mission risk, degraded safety margins, and limitations on where and how autonomous systems can be deployed. In high consequence applications such as logistics, inspection, emergency response, or defense, the inability to maintain trustworthy GPS based state estimation can impose operational constraints that prevent full autonomy from being realized.
03
Solution
Fusion Flight introduces a new class of GPS based sensing and inference technology that treats the received waveform as a rich, structured measurement source rather than a fragile positioning input. Instead of assuming that all distortions must be removed, our system learns to interpret the physical signatures embedded within multipath, interference, and atmospheric effects. By analyzing pre correlation and post correlation GPS data with physics guided machine learning models, the system identifies when a measurement is corrupted, characterizes the type of distortion, and quantifies its expected influence on the navigation solution. This transforms noise into information. Multipath becomes an indicator of environmental geometry, interference becomes a marker of risk, and subtle changes in the correlation function become cues that reveal the integrity of the signal path. The result is a navigation engine that understands the structure of the RF environment and remains self aware even when the incoming GPS stream violates classical assumptions.
The impact of this technology is substantial. By extracting actionable information from corrupted GPS signals, Fusion Flight enables autonomous UAVs and aerospace platforms to maintain reliable state estimation in environments that were previously inaccessible. Navigation integrity no longer collapses in dense urban corridors, under low altitude clutter, or in regions characterized by severe reflections. In mission critical settings, this capability preserves safety margins, reduces drift, and allows autonomous systems to operate with confidence without depending on ideal signal conditions. More broadly, the technology expands the operational envelope of autonomous flight. It supports persistent operation in contested RF environments, enhances reliability for long range missions, and provides a pathway for high density fleet coordination in complex airspace. As operations continue to shift toward environments where multipath dominates, the ability to interpret corrupted GPS signals will define the next generation of resilient autonomy.
04
Commercialization
The commercial trajectory of this technology has already gained significant momentum. Fusion Flight’s navigation resilience engine directly addresses a high priority gap in the rapidly expanding UAV and autonomous aviation markets, where operators routinely encounter degraded GNSS conditions that undermine mission reliability. The ability to extract meaningful environmental information from distorted GPS signals expands the operational envelope for commercial drones, urban air mobility vehicles, and autonomous inspection platforms. This creates measurable economic value in logistics, energy infrastructure monitoring, construction, maritime operations, and public safety, where navigation failures translate into missed missions, safety risks, and regulatory barriers. By enabling trustworthy operation in complex RF and multipath dominated environments, the technology positions itself as a foundational capability for any organization seeking to scale autonomous aerial operations.
We have successfully raised funding to transition this research toward commercialization through our startup, GUIDEAIR Labs. This support accelerates the maturation of the technology from a laboratory validated system into a deployable product that integrates with existing GNSS receivers, autopilots, and autonomy frameworks. Our commercialization plan includes structured hardware in the loop testing, integration with industry standard flight controllers, and joint evaluation pilots with early adopters in the aerospace and UAV sectors. GUIDEAIR Labs is building a scalable software and firmware stack that can be embedded into both new and legacy platforms, offering a practical upgrade path for fleet operators seeking higher reliability without replacing existing hardware. The funding we have secured enables focused engineering, regulatory alignment, and field validation efforts that will bring this capability to market as a trusted, commercially supported navigation resilience solution.
Publications
In progress
Reference
Add a Title
Reference
Ackhnowledgements
UW CoMotion
NSF
SAFRAN - MINERVA PROGRAM
UW ECE
WRF
Gallery
