ACAS X, the Collision Avoidance System Being Engineered to Replace TCAS, and What Dynamic Programming Changes About the Logic That Keeps Aircraft Apart

The aviation industry has spent over a decade quietly engineering a replacement for TCAS II, the collision avoidance system that has prevented countless midair collisions since the early 1990s. ACAS X - Airborne Collision Avoidance System, next generation - replaces fixed decision lookup tables with a Markov Decision Process, a probabilistic optimization framework that computes the best response across all probable encounter outcomes rather than matching scenarios to pre-written rules. The result is a system architecturally designed for an airspace that includes unmanned aircraft, electric VTOL vehicles, and traffic densities that simply did not exist when TCAS was engineered.

Why TCAS Exists and How It Works

By the late 1970s, the collision avoidance problem in commercial aviation was becoming acute. Traffic was increasing faster than radar coverage could keep pace, and a series of midair collisions and near-misses made clear that relying solely on see-and-avoid and controller intervention was not adequate in all conditions. The system needed a last layer of protection that lived on the aircraft itself.

TCAS works by putting a transponder interrogator on each aircraft. It pings nearby aircraft, listens for transponder replies, builds a geometric picture of surrounding traffic, and calculates whether any track is on a collision course. When a conflict is detected, it issues a Resolution Advisory - an RA.

The critical engineering challenge was coordination. If both aircraft climb to avoid the same conflict, the problem gets worse, not better. TCAS solves this by using a coordination channel on 1090 MHz to let the two threat aircraft communicate directly and assign complementary maneuvers. One climbs, one descends. The logic runs in roughly one second. Pilots are expected to respond within five seconds and reach the target vertical speed within three seconds after that.

After a midair collision over Cerritos, California in 1986 between a commercial airliner and a private aircraft, Congress accelerated the push for a mandate. TCAS II became mandatory for large air carriers by the early 1990s and spread internationally through ICAO. Over the system’s lifetime, some estimates credit it with saving thousands of lives.

Three Limitations That Justified a Replacement

TCAS II uses resolution advisory logic tables - an enormous decision matrix built by human engineers over years of analysis and simulation. It measures closing rate, time to closest approach (called tau), altitude, and the other aircraft’s current behavior, then looks up the prescribed response. This approach worked for thirty years. But its limitations have grown harder to ignore.

Nuisance alerts are the first problem. The logic tables are built to be conservative, so TCAS sometimes fires advisories in situations where a more sophisticated system would recognize the conflict will resolve without intervention. Nuisance alerts are not just an inconvenience - they erode pilot trust. Studies consistently document a cry-wolf effect: repeated false alarms reduce the urgency of response to real ones, which is exactly the wrong outcome for a safety-critical system.

Rigidity is the second problem. The logic tables were designed around a specific population of aircraft - transport-category jets and turboprops with known performance envelopes. When an unmanned aircraft with climb performance completely outside those assumptions enters the mix, the table has no entry for it. The same is true for electric VTOL vehicles with performance profiles that did not exist when the tables were written.

Static architecture is the third problem. The logic tables were computed once, certified, and deployed. Updating them requires a full new certification cycle. In an airspace that is about to change faster than any previous generation, a static decision engine is a structural vulnerability.

How ACAS X Thinks Differently: The Markov Decision Process

ACAS X, developed by MIT Lincoln Laboratory and the FAA beginning around 2008, addresses all three problems at the root.

The core insight: instead of a lookup table assembled by hand, use an optimization algorithm to compute the best action for every possible encounter state. The specific framework is a Markov Decision Process (MDP) - a method for making decisions under uncertainty when outcomes depend on both your action and states you cannot perfectly observe.

When ACAS X detects a potential threat, it does not ask whether a scenario matches a threshold. It asks: what action produces the best expected outcome, weighted across all the probable ways this encounter could play out? The system models the probability distribution of what might happen next, considers the full range of encounter developments, and calculates the expected cost of each possible action.

This means ACAS X holds its fire when the probability distribution says the conflict will likely resolve without intervention - eliminating many nuisance alerts. It fires earlier when the trajectory is clearly converging and waiting would cost response time. The alert fires with more confidence when it matters and more restraint when it does not.

The advisory logic was not programmed scenario by scenario. It was computed. Researchers defined the cost function - the mathematical expression of what outcomes are bad and how bad - defined probability models for how encounter geometries evolve, and ran the optimization. The output is still a table, but one generated by a principled optimization rather than assembled by committee. Critically, that optimization can be rerun when the operational environment changes. Add a new class of aircraft, update the probability models, recompute. The table reflects the new world without requiring engineers to manually re-examine thousands of scenarios.

The Four Variants of ACAS X

ACAS X is not a single product. It is a family of four variants designed for different operational contexts.

ACAS Xa is the direct replacement for TCAS II on large commercial aircraft. It uses the same transponder infrastructure, the same 1090 MHz frequencies, and the same pilot interface. The decision logic underneath is different. The goal is a substantial reduction in nuisance RAs while maintaining or improving performance against genuine conflicts.

ACAS Xo addresses specific operations where standard ACAS Xa behavior creates problems. The clearest example is closely-spaced parallel runway approaches. At airports like San Francisco International, simultaneous approaches to parallel runways are routine and safe, but TCAS sometimes fires because the two aircraft are intentionally flying in managed close proximity. ACAS Xo includes awareness of these environments and can suppress advisories that would fire under the standard algorithm but do not reflect actual risk in that context.

ACAS Xu is the variant with the most direct implications for the future of the airspace. The u stands for unmanned. This is the collision avoidance standard being built for UAS operating under Beyond Visual Line of Sight (BVLOS) rules in shared airspace with manned aircraft. For an unmanned system, the collision avoidance command goes directly to the autopilot - there is no pilot to receive a voice advisory. The logic has to work in that context, with different performance envelopes and different computational response latencies.

ACAS sXu is designed for small UAS - commercial delivery drones and inspection aircraft operating at lower altitudes in urban airspace. Same underlying framework, scaled for smaller, slower, lower aircraft.

Why This Matters for Pilots Flying TCAS Today

For crews flying TCAS II-equipped aircraft, the practical experience of an ACAS Xa RA will be similar to today. Voice alert. Vertical speed guidance. You respond. The human interface is not being redesigned.

What changes is the upstream logic deciding when to issue the advisory, what direction to assign, and how coordination between aircraft is handled. In practice, pilots should experience fewer events that feel like false alarms. When the system fires, it should feel more warranted.

What does not change is the fundamental rule: when an RA fires, you follow it. The coordination protocol between aircraft depends on both crews executing their assigned advisory. If one crew substitutes their own judgment or follows an ATC instruction instead of the RA, they break the protocol at the worst possible moment - while the other aircraft is already maneuvering on the assumption that you are executing yours.

Überlingen: The Accident That Makes This Principle Non-Negotiable

The worst TCAS-related accident in aviation history was not caused by a system failure. It was caused by a crew following an air traffic controller instead of their RA.

On July 1, 2002, a Bashkirian Airlines Tupolev and a DHL Boeing 757 collided over Überlingen, Germany, on Lake Constance. Seventy-one people died. TCAS worked correctly. Both aircraft received coordinated advisories - the Tupolev a descend command, the 757 a climb command. The DHL crew followed their RA and climbed. The Bashkirian crew received a conflicting instruction from their controller and followed the controller, descending into the climbing 757.

The investigation concluded unambiguously: the system performed as designed. Human deviation from the system’s coordinated advisory killed 71 people.

That lesson does not expire with a new system. ACAS X issues RAs for the same reason TCAS II did - the aircraft has better real-time collision geometry information than any ground controller. That principle is load-bearing regardless of what algorithm is running underneath.

Where ACAS X Stands in Development

The core ACAS X algorithms have been published in technical literature and are publicly available. MIT Lincoln Laboratory has been transparent about the methodology. RTCA - formerly the Radio Technical Commission for Aeronautics - produced standards framework document DO-365 for ACAS X systems. The FAA has conducted encounter simulations running into the hundreds of millions of synthetic conflict scenarios, testing the new logic against historical traffic data and against conditions that do not yet exist in the current airspace.

As of late 2026, firm timelines for a mandate replacing TCAS II with ACAS Xa are not locked in. Aviation mandates take time, and they should - the certification process for avionics hardware, coordination with international regulators through ICAO, and fleet transition across thousands of aircraft are not quick processes. Some technical discussions place a possible mandate timeline in the early 2030s.

ACAS Xu is in some ways on a faster track, because there is no legacy fleet to retrofit. New UAS platforms being designed for BVLOS operations are incorporating collision avoidance from the start, and ACAS Xu is the standard they are building toward.

The Bigger Picture: Engineering for the Airspace Ahead

The airspace being built over the next two decades is fundamentally different from the one TCAS was designed for. Thousands of commercial UAS operating BVLOS. Advanced air mobility vehicles with vertical performance profiles that have no analogue in existing tables. Autonomous aircraft with no pilot to receive a voice advisory. Urban corridors operating at densities and altitudes that did not exist in the 1990s.

A table-based system built for a specific, known population of aircraft cannot evolve to handle that at any acceptable pace. A probabilistic optimization framework that can be recomputed when the environment changes, can. That is why ACAS X is not simply an upgrade to TCAS - it is a rearchitecting of how the airspace manages conflict at the aircraft level, while most of the fleet is still running logic tables encoded when the first generation of glass cockpits was appearing on the ramp.

Key Takeaways

TCAS II uses static logic tables built by engineers to match encounter scenarios to prescribed advisories; the tables cannot be updated without a full certification cycle.
ACAS X uses a Markov Decision Process to compute the optimal action across all probable encounter outcomes, reducing nuisance alerts and adapting to new aircraft types by rerunning the optimization.
Four variants cover the full future airspace: ACAS Xa for commercial jets, ACAS Xo for parallel runway environments, ACAS Xu for BVLOS unmanned aircraft, and ACAS sXu for small UAS.
A mandate for ACAS Xa is not finalized as of late 2026; technical discussions suggest the early 2030s, with ACAS Xu on a faster track due to the absence of a legacy fleet.
The RA compliance rule does not change: when a resolution advisory fires, follow it immediately - the 2002 Überlingen collision that killed 71 people was caused by a crew following ATC instead of their RA while the other aircraft was already maneuvering on the coordinated advisory.