Beyond Visual Line of SightThe Layered Safety Architecture of UAV Collision Avoidance

1The Foundational Safety Architecture of Conventional Aviation

The safe operation of aircraft in modern, complex airspace rests on a deeply layered and redundant safety architecture built up over decades. That framework combines strategic separation, tactical surveillance, and last-resort avoidance, and it forms the baseline against which any UAV collision avoidance capability must be measured. Its cornerstone is "see and avoid" — traditionally a human pilot visually scanning for other aircraft. For UAVs, which have no pilot on board, replicating this is not a matter of installing a camera; it requires an entire autonomous perception and decision layer that fulfills the same safety intent.

The primary defense is Air Traffic Control (ATC) and the wider air traffic management system. Flight planning establishes strategic separation before takeoff by assigning routes, altitudes, and time slots. In controlled airspace, aircraft flying under Instrument Flight Rules are subject to ATC authority, which uses radar and other surveillance tools to maintain minimum distances. Regulations such as 14 CFR Part 91 supply fallback right-of-way rules — for example, requiring both aircraft in a head-on encounter to alter course to the right, or requiring an overtaking aircraft to give way — but these depend on pilots actually being able to see one another.

Modern ATC surveillance is built on cooperative technology. Mode S transponders respond to radar interrogation with an aircraft's identity and other data, and the more recent Automatic Dependent Surveillance–Broadcast (ADS-B) has aircraft broadcast their own GNSS-derived position, altitude, and velocity. ADS-B Out feeds this to ground stations and equipped aircraft, sharpening situational awareness for both controllers and pilots — but the whole system depends on the target aircraft actually carrying functioning, correctly configured equipment.

As a final, independent safety net, Airborne Collision Avoidance Systems (ACAS) — most commonly the Traffic Alert and Collision Avoidance System (TCAS) — operate without relying on ATC instructions at all. TCAS interrogates nearby transponders to calculate range, bearing, and altitude; if a conflict develops it first issues a Traffic Advisory, and if the threat continues to close, a Resolution Advisory with explicit vertical guidance such as "Climb, Climb" or "Descend, Descend." RAs are designed to resolve the conflict even if that means deviating from a current ATC clearance. The next generation, ACAS X, is being tuned to provide Remain Well Clear alerting aligned with the newer unmanned-aircraft standards, narrowing the gap between crewed and uncrewed collision-avoidance philosophy. From strategic flight planning through tactical ATC commands to TCAS's autonomous last line of defense, the entire architecture assumes aircraft can either be seen or can broadcast their state so others can see them — and it is precisely that assumption that a pilotless aircraft has to solve for.

2Defining Detect-and-Avoid: From Concept to Certified System

Unmanned Aircraft Systems (UAS) represent a fundamental departure from this architecture because there is no onboard pilot to perform the visual scan. Detect-and-Avoid (DAA) is the aviation community's answer: not a single piece of hardware but a system of sensors, processing algorithms, and pilot interfaces that let a remotely piloted aircraft sense, assess, and act on nearby traffic, fulfilling the operational intent of "see and avoid" from a Ground Control Station. Its functional chain mirrors a human pilot's own cognitive process: Detect, Track, Predict, Evaluate, Remain Well Clear (RWC), and finally Collision Avoidance (CA).

Detect identifies potential traffic via sensors — an ADS-B reception, or radar reflecting off a non-cooperative object. Track builds a consistent state vector for that object over time, estimating position, velocity, and heading. Predict projects both the UAV's and the intruder's future trajectories. Evaluate continuously checks whether those projected paths will breach the separation minima. If a conflict is caught early, Remain Well Clear generates alerts or maneuver guidance so the remote pilot can preserve safe distance before things get critical — its threshold is a defined regulatory parameter, not a rule of thumb. If separation keeps deteriorating despite RWC action, the final layer, Collision Avoidance, engages with more forceful, last-resort guidance.

RWC and CA are deliberately not synonymous: RWC is tactical separation management, while CA is an emergency procedure, and the logic for transitioning between them is carefully bounded so CA is only invoked when truly necessary. The specific "well clear" thresholds are not arbitrary — they come out of risk analysis that accounts for airspace class, the UAV's performance characteristics, encounter geometry, and the terms of a given operating approval, with research exploring both horizontal miss distance and time-based separation rather than one universal number. Standards bodies like RTCA and EUROCAE codify the minimum performance for each stage of this chain, so that safety, reliability, and interoperability hold across manufacturers and platforms.

3Cooperative Surveillance and Non-Cooperative Detection

A central design challenge is the split between cooperative and non-cooperative traffic, because it dictates which sensing technology a DAA system needs and how complex the resulting architecture becomes. A cooperative target actively broadcasts its state — via Mode S transponder or ADS-B. A non-cooperative target either lacks the equipment, has it malfunctioning, or is otherwise not broadcasting usable data. A DAA system that only listens for cooperative traffic is inherently incomplete: it is blind to anything not "talking" to it.

ADS-B In gives a UAV's DAA system direct access to a cooperative aircraft's GNSS-derived position, altitude, velocity, and identity, which sharply reduces the sensor-fusion workload needed just to establish a track. But it is not a complete solution: older aircraft, some military platforms, and general aviation aircraft may not carry ADS-B, may have it switched off, or may suffer a transponder or GNSS failure that turns them non-cooperative regardless of their original equipment.

Independent sensing fills that gap. Air-to-air radar generates its own energy, so it works regardless of what the target aircraft is carrying — making it essential for non-cooperative detection, with modern systems increasingly using Active Electronically Scanned Array (AESA) radar for better performance in a smaller package. Electro-Optical/Infrared (EO/IR) imaging complements this with passive detection: EO captures visible light, IR captures heat, and computer vision can automatically detect, classify, and track what the cameras see — without the UAV emitting anything that could reveal its own position. The tradeoff is that EO/IR is heavily degraded by cloud, fog, rain, and darkness. A third approach, Ground-Based Detect-and-Avoid (GBDAA), places radar or EO/IR sensor towers on the ground to monitor a defined volume of airspace and relay the resulting traffic picture to the UAV or its ground station — uAvionix's Casia G is a working example, and it was instrumental in an FAA BVLOS waiver for Intellition. No single technology guarantees detection in every scenario, which is why a genuinely comprehensive DAA system fuses cooperative data with independent radar, EO/IR, and/or ground-based inputs into one resilient picture.

4Sensor Fusion: Why No Single Sensor Is Enough

Modern DAA architectures are built on the principle of sensor fusion — combining multiple complementary sensors into a picture more accurate and robust than any one of them could produce alone. This mitigates each technology's weaknesses while keeping its strengths, and it is essential for meeting civil aviation authorities' safety bar. A fused system typically starts with ADS-B for cooperative traffic, since it offers precise, GNSS-derived state data and a reliable base for initial tracks — but it leaves a blind spot wherever non-cooperative traffic is flying.

Airborne radar closes that gap by detecting objects regardless of cooperative status, and correlating a radar track against an ADS-B track sharing the same identity and motion both confirms the cooperative target and adds redundancy: if an ADS-B signal were ever spoofed or jammed, the radar track would still hold. EO/IR sensors add a passive layer on top, useful for resolving ambiguity — a radar return might be visually confirmed by EO/IR as a small general-aviation aircraft, helping a remote pilot judge the actual risk. Fusing these streams requires real data-association and track-correlation algorithms so that the picture handed to the pilot or the flight-control system is decluttered and reliable rather than a pile of unreconciled detections.

Ground-Based Detect-and-Avoid (GBDAA) adds a further layer by moving heavy sensing and processing to fixed or mobile ground stations, which reduces the weight, power, and cooling burden on the aircraft itself and allows for more powerful ground-side computing than a mass-constrained platform could carry. uAvionix's Casia system, using ground-based computer vision for passive non-cooperative detection, has already been accepted by the FAA as part of a BVLOS waiver package. In a fully integrated architecture, GBDAA output feeds the same fusion processor as the onboard sensors, extending the aircraft's effective awareness beyond its own immediate field of regard.

5The Regulatory Framework Enabling Safe Integration

Building DAA systems is not purely an engineering exercise — it is inseparable from an evolving global regulatory framework, shaped by ICAO at the international level and by regional authorities such as the FAA and EASA. Specialized technical committees — RTCA Special Committee 228 in the U.S. and EUROCAE Working Group 105 in Europe — develop the Minimum Operational Performance Standards (MOPS) that define what a DAA system must actually do.

RTCA DO-365 is the seminal MOPS document for DAA systems on large unmanned aircraft in controlled airspace, specifying traffic detection, conflict prediction, and RWC/CA alerting along with concrete performance metrics like detection range and update rate. RTCA DO-366 does the same for Air-to-Air Radar (ATAR), a core DAA sensor. These MOPS translate high-level safety goals into verifiable technical specifications without over-prescribing implementation, leaving room for innovation on top of a guaranteed safety floor.

The FAA turns MOPS into binding Technical Standard Orders: TSO-C211a for DAA systems and TSO-C212a for ATAR. A manufacturer must obtain a TSO Authorization or Letter of Design Approval demonstrating compliance before a DAA system can move toward type certification on an aircraft — an independent verification step for every certified system in the U.S. National Airspace System. Europe is pursuing parallel standards, including EASA/EUROCAE's ED-258 for DAA MOPS and ED-264 for CA system interoperability, with harmonization between FAA and EASA regulation a stated goal for streamlining global certification.

At the international level, ICAO has amended Annex 6 of the Chicago Convention with a new Part IV specifically for Remotely Piloted Aircraft Systems, establishing foundational standards for operator certification and requiring a Detect-and-Avoid capability for international RPAS flights, effective from late 2026. On the military side, NATO has promulgated its own "Sense and Avoid for Unmanned Aircraft Systems" standard — composed of AEP-107 and STANAG 4811 — to standardize minimum functional and performance requirements across the Alliance and ensure compatibility with civil operations. Together, these layers of national, regional, and international rulemaking form the scaffolding that will let UAVs share the sky with passenger jets, cargo aircraft, and helicopters on a routine basis.

6Scenarios in Shared Airspace: Civil and Military Integration

Consider two large, civil-certified UAVs flying BVLOS on intersecting Class B routes. Each aircraft's ADS-B In receiver picks up the other's cooperative transmission while its airborne radar sweeps for anything non-cooperative; the fusion tracker combines both inputs and predictive algorithms calculate that the projected paths will breach separation minima in roughly three minutes. That triggers Remain Well Clear: an advisory suggesting a coordinated altitude change or slight course deviation, which the remote pilots execute to pass well clear of one another — while ATC maintains strategic separation throughout and the tactical deconfliction runs autonomously through the DAA systems. A UTM or U-space traffic-management layer could resolve the same conflict even earlier, by assigning routes and time windows before it ever arises.

A more complex case: a medium-altitude, long-endurance (MALE) UAV in a controlled corridor encounters a civil airliner on a crossing IFR flight plan. ATC's altitude and sequencing assignments are the first line of separation. If that procedural separation began to break down, the UAV's multi-sensor DAA suite — ADS-B, radar, possibly TCAS — would detect the approaching airliner. Caught early, RWC guides a corrective maneuver; if the conflict develops rapidly, the system escalates to Collision Avoidance, issuing a TCAS-like Resolution Advisory that commands a specific evasive maneuver even if it means overriding a conflicting ATC clearance. This is the clearest illustration of DAA as an independent safety layer that complements ATC rather than replacing it.

Military integration adds its own wrinkles. A U.S. Air Force MQ-9 Reaper has demonstrated safe NAS operation using a civilian-approved surveillance network, proving even large military platforms can follow civil procedure. The MQ-9B SkyGuardian, built to meet NATO STANAG 4671, is a step toward a genuinely "type-certifiable" unmanned aircraft that can integrate seamlessly with civil and allied traffic, using a multi-sensor DAA fusion of radar and other technologies to fulfill see-and-avoid in a manner analogous to a crewed aircraft. Military operations can still involve higher speeds, sharper maneuvers, and formation flight that differ from standard civil patterns, and DAA logic has to keep resolving conflicts against the same fundamental separation rules even as mission context shifts how a maneuver gets prioritized. NATO's Sense and Avoid standardization exists precisely to keep military UAVs compatible with non-segregated airspace and civil traffic, preventing the kind of dangerous misunderstanding that inconsistent procedures would otherwise invite. Civil or military, the underlying principle holds: safe, efficient mixed-use airspace depends on a layered system of separation services, surveillance, and autonomous DAA logic working together, not any one of them alone.

Sources