ACAS X, the MIT Algorithm Replacing TCAS, and the Collision Avoidance Logic That Pilots Need to Understand Before the Mandate Arrives

The Traffic Collision Avoidance System has protected aircraft since the 1980s and has effectively eliminated midair collisions between transport aircraft in the United States since becoming mandatory in the early 1990s. But the airspace TCAS was designed for no longer exists. A next-generation replacement called ACAS X, developed primarily at MIT Lincoln Laboratory in partnership with the FAA and Stanford researchers, is moving through international certification now - and every pilot who has ever responded to a resolution advisory needs to understand what is changing and what is not.

Why TCAS Works - and What It Cannot Handle

TCAS II, the version carried by nearly all transport category aircraft and many GA aircraft today, operates on a deterministic logic table. The system measures closure rate, altitude separation, and time to collision, then looks up a pre-defined answer. Given specific inputs, the output is always the same: climb, descend, or monitor.

The coordination between two TCAS-equipped aircraft is the engineering achievement that made the system work. Through a short-range data link, the two transponders negotiate directly, independent of ATC. One climbs, one descends, and the resolution advisories complement each other. ATC does not authorize the RA. ATC cannot stop it. You follow the RA first, advise ATC second - that is the correct order of operations and it is regulatory.

The record speaks for itself. Midair collisions between transport aircraft have been essentially eliminated in the United States since the TCAS II mandate. That is not coincidence.

Why TCAS Is Being Replaced

The airspace of the future is fundamentally different from the one TCAS was built for.

Unmanned aerial systems are operating at altitudes once exclusive to manned aviation. The FAA is advancing its Beyond Visual Line of Sight (BVLOS) frameworks. Electric vertical takeoff and landing (eVTOL) aircraft are beginning FAA certification processes now, some of them fully autonomous. When an autonomous aircraft receives a resolution advisory, no pilot responds - the flight management software does. What logic governs that response matters enormously.

There is also a documented operational problem with the current system: nuisance advisories. TCAS II works from geometry and closure rate, not a probabilistic assessment of actual threat. In high-density terminal environments, it generates RAs in situations where collision risk is genuinely low, because the lookup table cannot distinguish between a threatening encounter and a non-threatening one with similar geometry. Studies have documented nuisance advisory rates in certain terminal environments that are high enough to measurably impact system efficiency - every RA disrupts the flight, disrupts ATC sequencing, and in busy airspace those disruptions compound.

The goal is not to fix TCAS for the airspace we have. It is to build a system for the airspace we are going to have.

How ACAS X Works: Dynamic Programming Replaces the Logic Table

The Airborne Collision Avoidance System (ACAS) has existed in various forms for decades. The X designation marks the next generation. The fundamental change is not the sensors or the data link - it is the algorithm.

TCAS uses a deterministic lookup table. ACAS X uses dynamic programming, specifically a technique called a Markov Decision Process.

In practical terms: rather than looking up an answer, the system continuously calculates one. It builds a probabilistic model of how the encounter is likely to evolve, accounts for uncertainty in sensor data, and computes the action that produces the best expected outcome across the full range of possible futures - not just the worst case, but the entire probability distribution.

The system weighs the cost of a collision against the cost of a nuisance advisory and generates the advisory that best minimizes expected risk. It is a fundamentally different kind of math.

The results in simulation are significant. Researchers at MIT Lincoln Laboratory have published data showing ACAS X produces 40 to 50 percent fewer nuisance advisories than TCAS II in certain scenarios while maintaining or improving collision risk reduction. That is not a marginal refinement.

Dynamic programming also enables something a logic table cannot easily produce: graduated advisories. TCAS II is largely binary - climb or descend. ACAS X can generate more precise directives calibrated to the actual threat level, because it is calculating that level dynamically rather than reading it from a table.

The ACAS X Variants: A Family, Not a Single Product

ACAS X is not one system. It is a family of variants, each optimized for a different operational environment.

ACAS Xa is the successor to TCAS II - designed for commercial transport operations and the variant that will eventually appear in airline and turboprop cockpits. Both the FAA and the International Civil Aviation Organization (ICAO) are involved in transition planning, and ICAO has published standards work toward this variant.

ACAS Xo is tuned for oceanic operations, where traffic density is lower but the stakes of a missed encounter are equally high. The advisory logic is calibrated differently for that environment.

ACAS Xu is designed for large unmanned aircraft. The logic accounts for the different response characteristics of automated systems - when an unmanned aircraft receives an advisory, the response goes to the flight management system, not a human pilot.

ACAS sXu addresses small unmanned systems - the drone world. This is technically the hardest variant. Small unmanned aircraft operate at altitudes that create complex encounter geometry, may lack transponders entirely, and carry sensors constrained by size and power. Sense-and-avoid for the low-altitude UAS environment, where drones and eventual eVTOLs will mix with manned GA traffic, is one of the genuinely hard unsolved problems in current aviation engineering.

For GA pilots, ACAS Xa is the version that matters most directly.

What Changes for Pilots - and What Does Not

The most important rule does not change: when you receive an RA, you follow it immediately.

No deliberation. No checking with ATC first. No evaluating whether you agree. The coordinated avoidance logic assumes you are making your move while the other aircraft makes the complementary one. If you hesitate, the math breaks - and that breaks the protection the system provides.

What may change is the nature of the advisory itself. If ACAS Xa delivers a more graduated directive rather than the unambiguous “climb climb climb” call-out, the required response is still immediate compliance. Pilot training and FAA advisory circular guidance will need to be updated as the system rolls out to reflect the new advisory types.

This is not a trivial concern. When TCAS was first introduced, there was a documented pattern of pilots being reluctant to follow RAs that conflicted with ATC instructions or their own situational assessment. Years of training and several accidents reinforced the compliance requirement, and that hesitation largely disappeared. As ACAS X produces more sophisticated advisories - some of which will feel less urgent than the current call-out - there is a real question about whether human response will remain as immediate. Aviation’s history with automation shows that as cockpit systems become more nuanced, pilots tend to engage with them more as advisors and less as systems requiring immediate compliance. That tendency has caused accidents.

Training organizations and individual pilots will need to treat this actively, not assume the behavior carries over automatically.

The Limits That Do Not Change Regardless of the Algorithm

TCAS II and ACAS X both depend on Mode C and Mode S transponder replies. Traffic without working transponders is invisible. Traffic with degraded sensors presents corrupted geometry. ACAS X’s probabilistic framework handles uncertainty better than a lookup table - but it cannot make good decisions from bad data.

Visual scanning, an accurate mental model of what the traffic display actually shows, and treating electronic traffic information as one layer in a multi-layer defense remain requirements. The algorithm improving does not change the data quality problem.

Timeline and Where Things Stand

There is no firm mandatory date for the ACAS Xa transition. The standards work is ongoing through the FAA and ICAO processes, the system has been tested extensively in simulation and actual flight, and the direction of travel is clear. Published technical reports from MIT Lincoln Laboratory are publicly available - not classified documents - for pilots and engineers who want to go deeper into the underlying research.

The certification path is moving. The mandate will arrive. The question is when, not whether.

Key Takeaways

TCAS II uses a fixed logic table to generate resolution advisories; it has effectively eliminated transport aircraft midair collisions since its mandate in the early 1990s, but the system was not designed for unmanned aircraft, eVTOLs, or dense mixed-use airspace.
ACAS X uses a Markov Decision Process - a probabilistic calculation that continuously models the encounter and weighs collision risk against nuisance advisory cost, generating the optimal advisory across the full range of possible outcomes.
Nuisance advisory rates drop 40–50 percent in simulation under ACAS X compared to TCAS II, with maintained or improved collision risk reduction.
ACAS X is a family: ACAS Xa (commercial transport, the TCAS II successor), ACAS Xo (oceanic), ACAS Xu (large UAS), and ACAS sXu (small UAS).
The core compliance rule does not change: follow the RA immediately. As advisories become more graduated, maintaining that immediate-compliance discipline will require deliberate attention from training organizations and pilots.