Traffic_collision_avoidance_system

# Traffic collision avoidance system > [[PORTAL_Aviation|Aviation]] · [[PORTAL_Avionics|Avionics]] spine.  ## Microsims — three.js ### Traffic Alert and Collision Avoidance System (three.js) <div class="microsim-player"> <iframe src="https://wikitube-3d-microsims.netlify.app/flight/Traffic_collision_avoidance_system.html" width="100%" height="620" frameborder="0" loading="lazy" sandbox="allow-scripts allow-same-origin" title="Traffic Alert and Collision Avoidance System — three.js microsim"></iframe> </div> **Open it full-screen:** [flight/Traffic_collision_avoidance_system.html](https://wikitube-3d-microsims.netlify.app/flight/Traffic_collision_avoidance_system.html) · library `threejs` · route `microsim/threejs/` ### Related microsims Live sims on neighbouring articles: - [[Air_traffic_control]] - [[Transponder]] - [[Radar]] - [[Terrain_awareness_and_warning_system]] - [[Autopilot]] - [[Avionics]] *Sim hosted off-article; the article owns the reference, not the runtime (WIKI_RULES §10.4).*  ## Overview The **traffic alert and collision avoidance system** (TCAS; **ACAS II** in ICAO usage, where ACAS stands for airborne collision avoidance system) is an aircraft-carried system that interrogates the transponders of nearby aircraft, tracks them, and warns the crew when one of them is going to hit them. It is entirely independent of the ground. It knows nothing about flight levels, sectors, clearances, airspace classes or who is talking to whom. It has one job, and it does that job in the last minute of an encounter. It is worth stating at the outset what TCAS is **not**. It is not a separation service. Separation is provided by [[Air_traffic_control]] and is measured in miles and thousands of feet across horizons of minutes. TCAS acts tens of seconds before impact and settles for a few hundred feet of vertical miss. By the time it issues a resolution advisory, the separation minima have already been lost or are seconds from being lost. ICAO, EUROCONTROL and the FAA are consistent and emphatic on this point: ACAS is a **safety net of last resort**, its performance is not to be relied upon in planning, and using it to space aircraft — flying by the traffic display, in particular — is explicitly prohibited. A system that alerts perhaps once in tens of thousands of flight hours is not a service; it is an airbag. Three generations exist. **TCAS I** issues traffic advisories only and is fitted to smaller commuter aircraft. **TCAS II** adds **resolution advisories** — vertical avoidance manoeuvres — and is what the word normally means. ICAO Annex 6 requires ACAS II on turbine aeroplanes above 5700 kg maximum take-off mass or authorised to carry more than 19 passengers. **TCAS III**, which was to have added horizontal resolution advisories, was abandoned: the bearing measurement from a small directional antenna is good to roughly ten degrees, nowhere near accurate enough to promise a horizontal miss. That single hardware limitation is why every TCAS II advisory to this day is purely vertical. The successor now entering service, **ACAS X**, replaces the hand-tuned threshold logic with a numerically optimised lookup table. The version in service since the early 2010s is **TCAS II version 7.1**. It differs from version 7.0 in two changes, both small in code and large in consequence, and both traceable to a single accident: the mid-air collision over Überlingen, Germany, on 1 July 2002, in which 71 people died. The microsim above puts two — and then up to six — aircraft on converging tracks and shows the quantity that actually drives the alerting: **tau**, the estimated time remaining to the closest point of approach. Each intruder's ground track is ruled not in miles but in **seconds of tau**, with the traffic-advisory and resolution-advisory bands marked. Change the closure speed and those coloured bands physically stretch and shrink in miles while staying exactly the same length in seconds. Then turn the coordination link off and watch the protection collapse. ## The physics ### Range from a round trip TCAS is an interrogator. It transmits on 1030 MHz and listens on 1090 MHz — the same pair of frequencies used by secondary surveillance [[Radar]] — and every aircraft carrying a [[Transponder]] answers. Mode C transponders are interrogated with a broad "whisper–shout" sequence of stepped power levels, a technique that limits how many aircraft reply at once and so limits mutual garbling. Mode S transponders are interrogated selectively, by their unique 24-bit ICAO aircraft address, which is far more efficient and is what makes coordination possible at all. Range comes from the round trip. The transponder's turn-round delay is fixed by standard — nominally 3.0 μs for Mode A/C and 128 μs for Mode S — so $ r = \frac{c\,(\Delta t - t_{\text{reply}})}{2} $ gives slant range to an accuracy of a few tens of feet. Altitude comes from the reply itself: 100 ft quantisation from a classic Mode C transponder, 25 ft from Mode S. Bearing comes from a four-element directional antenna and is poor. So TCAS knows range well, altitude reasonably, and direction badly — which is exactly the information profile that produces a vertical-only collision avoidance system. An intruder that does not report altitude can be given a traffic advisory but never a resolution advisory: without altitude there is no way to know whether a manoeuvre helps or hurts. ### Tau, and why time is the right variable Given range and its rate of change, define $ \tau = -\frac{r}{\dot r} $ which is the time to zero range if both aircraft continue as they are. Everything TCAS does is keyed to this number, and the reason is easiest to see from two examples. Two aircraft 3 NM apart, closing at 20 kt, have $\tau = 540$ s. They are nine minutes from each other. Nothing should happen, and nothing does. Two aircraft 8 NM apart, closing at 1000 kt — an ordinary head-on pair of jets — have $\tau = 28.8$ s. They are on the edge of a collision. **A distance test cannot tell those two situations apart.** Any threshold that alerts on the second must alert on the first, and a system that alerts on the first is a system crews will switch off. The consequence is that the geometry of the alert is elastic. At sensitivity level 7, the resolution-advisory threshold is 35 s. Thirty-five seconds is 11.7 NM at 1200 kt of closure, 8.0 NM at 820 kt, 4.9 NM at 500 kt, and less than a mile at 100 kt. The microsim draws this directly: the amber and red bands on the intruder's track are fixed in seconds and visibly change length in miles as the closure-speed control moves. A slow convergence at close range can be perfectly legal; a fast one at long range is not. ### Modified tau and the DMOD floor Range tau has a defect. As $\dot r \to 0$ it goes to infinity, so two aircraft drifting together at walking pace from half a mile apart would never trip the test at all. TCAS therefore uses **modified tau**, which puts a distance floor underneath: $ \tau_{\text{mod}} = -\frac{r^2 - \text{DMOD}^2}{r\,\dot r} $ DMOD is the distance modification, a threshold-specific constant. When $r > \text{DMOD}$ this behaves much like range tau; when $r \le \text{DMOD}$ the numerator goes non-positive and the test trips immediately whatever the closure rate. At sensitivity level 7, DMOD for a resolution advisory is 1.10 NM. So in the 100 kt example above, where 35 s of tau corresponds to only 0.97 NM, it is the DMOD floor and not the tau threshold that fires first. This is not a patch; it is the recognition that "time to go" is a bad estimator of danger when nothing much is happening quickly. ### The vertical test The range test alone is not sufficient. An aircraft 5000 ft below is no threat however fast it is closing horizontally. Every advisory therefore requires a **second, independent test on the vertical dimension**, and both must pass: $ |\Delta z| \le \text{ZTHR} \quad\text{or}\quad \tau_v = -\frac{\Delta z}{\Delta \dot z} \le \tau_{\text{threshold}} $ ZTHR is the vertical threshold, 600–800 ft for an RA depending on sensitivity level. The vertical tau is the time until the two altitudes cross, if they are converging at all. An intruder that is level 1500 ft below and staying there fails the vertical test permanently and can never generate an RA, no matter how close it comes horizontally. ### Sensitivity levels TCAS is not one system with one threshold. It is seven, selected automatically by own altitude, because the correct amount of warning is not a constant. Low down, closure rates are small and traffic is dense, and a generous threshold would fire on every circuit; high up, closure rates are large and the aeroplane is sluggish in the vertical, and the same system needs the better part of a minute. | SL | Own altitude | TA τ (s) | RA τ (s) | DMOD TA / RA (NM) | ZTHR TA / RA (ft) | ALIM (ft) | | --- | --- | --- | --- | --- | --- | --- | | 2 | below 1000 ft AGL | 20 | *RAs inhibited* | 0.30 / — | 850 / — | — | | 3 | 1000–2350 ft AGL | 25 | 15 | 0.33 / 0.20 | 850 / 600 | 300 | | 4 | 2350 ft AGL – 5000 ft | 30 | 20 | 0.48 / 0.35 | 850 / 600 | 300 | | 5 | 5000–10 000 ft | 40 | 25 | 0.75 / 0.55 | 850 / 600 | 350 | | 6 | 10 000–20 000 ft | 45 | 30 | 1.00 / 0.80 | 850 / 600 | 400 | | 7 | 20 000–42 000 ft | 48 | 35 | 1.30 / 1.10 | 850 / 700 | 600 | | 7 | above 42 000 ft | 48 | 35 | 1.30 / 1.10 | 1200 / 800 | 700 | Quoting "TCAS alerts at 25 seconds" or "at 40 seconds" as though it were a universal number is therefore simply wrong. The honest statement is that a traffic advisory comes at roughly **40 to 48 seconds** of tau in the altitude bands where airliners cruise and climb, and a resolution advisory at roughly **20 to 35 seconds**, with both figures falling steeply towards the ground. Below 1000 ft above ground level, resolution advisories are inhibited altogether; "Descend" is inhibited below about 1100 ft AGL and "Increase descent" below about 1450 ft; traffic advisories are inhibited below about 500 ft. TCAS knows that near the ground the ground is the bigger problem — which is where the [[Terrain_awareness_and_warning_system]] takes over, and why a TAWS warning takes priority over a TCAS RA. The last column, **ALIM**, is the altitude limit: the vertical miss distance an RA is trying to achieve at the closest point of approach. It is 300 ft down low and 600–700 ft at cruise. That is the whole ambition of the system. Set against the 1000 ft vertical separation minimum that air traffic control works to, it is a reminder of where in the sequence of defences TCAS actually sits. ### Choosing a sense Once an RA is warranted, TCAS must choose a **sense**: climb or descend. It does this by simulation. It models what would happen if it climbed and what would happen if it descended, using an explicit assumption about how the crew will respond — **5 seconds of nothing, then 0.25 g** for an initial advisory, and **2.5 seconds, then 0.35 g** for a subsequent one — and projects the vertical separation at the closest point of approach in each case. It selects the sense that achieves ALIM. If both do, it prefers the **non-crossing** sense: the one that does not require passing through the intruder's altitude. The strength is chosen the same way. If the aircraft is already doing something that will clear ALIM without any change, the RA is **preventive**: "Monitor vertical speed", meaning do not enter the restricted vertical band. Otherwise it is **corrective**, and the initial demand is 1500 ft/min. If that turns out to be insufficient, "Increase climb" or "Increase descent" raises it to 2500 ft/min. Now consider the case where the two aircraft are at the **same altitude**, which is not an exotic case — it is what two aircraft on the same cleared flight level look like. There is no non-crossing sense. The two projections are mirror images. The comparison is a dead heat, and the tie has to be resolved by some rule inside the box. Both boxes hold the same rule. **Both boxes therefore return the same answer.** That is not a bug and it is not bad luck; it is what identical deterministic algorithms do when fed identical inputs. ### Coordination over Mode S This is the reason the Mode S data link is in the loop, and it is the centrepiece of the whole design. When both aircraft carry TCAS II, each sends the other **coordination interrogations** once per second, carrying its **vertical resolution advisory complement** — in effect, "I have taken the up sense; you are constrained to down". The receiving unit is then not free to choose. The two senses are complementary by protocol, not by luck. If both aircraft happen to select on the same one-second cycle, the tie is broken deterministically on the **24-bit ICAO aircraft address**, so that exactly one of the two yields and the outcome is never symmetric. Some documentation describes the tie in terms of the higher-addressed aircraft being the one that changes; the load-bearing fact is not which way round it goes but that the rule is fixed, shared and asymmetric. Symmetry has to be broken by something, and an arbitrary but agreed number is the cheapest something available. Turn the coordination control in the microsim to **OFF** with the vertical offset at zero, and watch what the argument above predicts: both units choose climb, both aeroplanes climb, the projected vertical miss at the closest point of approach stays at zero, and the encounter ends in a near mid-air collision. Then set the vertical offset to 400 ft and turn coordination off again — and the senses come out complementary anyway, because now each unit can see who is above whom and the non-crossing rule does the work. That is the honest shape of the result: **coordination matters most in exactly the symmetric geometry that a shared cleared flight level produces**, which is to say, in the common case. ### When ATC and the RA disagree A controller cannot normally see that an RA has been issued. (Some ground systems now display downlinked RA reports, which Mode S makes possible, but this is far from universal and arrived long after the logic did.) So a controller may issue an instruction that points the opposite way to the advisory the crew has just received. The rule is unambiguous and has been since 2003: > **The crew follows the resolution advisory.** ICAO's ACAS operating procedures in PANS-OPS (Doc 8168) require pilots to respond immediately to an RA, never to manoeuvre in the opposite sense to one, and explicitly to follow the RA **even if it conflicts with an air traffic control instruction to manoeuvre**. The one stated exception is where compliance would itself jeopardise the safety of the aeroplane, which is the pilot-in-command's final authority reasserting itself at the last possible moment. The crew notifies ATC as soon as workload permits and returns to the clearance when the conflict is resolved. PANS-ATM (Doc 4444) closes the loop from the ground: once an aircraft departs from its clearance in compliance with an RA, the controller **ceases to be responsible** for separating it from the traffic concerned, and must not attempt to modify its flight path until the crew reports returning to the clearance. In the United States 14 CFR 91.123(a) writes the same priority into the regulations. The microsim lets you create the conflict directly. **Issue contradictory ATC instruction** gives the intruder a controller's instruction pointing the opposite way to its RA; both commands are then drawn, the green RA arrow and the magenta ATC arrow, pointing in opposite directions. By default the crew flies the RA, and the encounter resolves. Tick **Crew flies the ATC instruction instead of the RA** and it does not. ### Überlingen, 1 July 2002, and version 7.1 At 21:35 UTC on 1 July 2002, near Überlingen on Lake Constance, a Bashkirian Airlines Tupolev Tu-154M (flight 2937, Moscow to Barcelona) and a DHL Boeing 757-200F (flight 611, Bergamo to Brussels) collided at about 34 900 ft. All 71 people aboard both aircraft were killed: 69 on the Tupolev, of whom 45 were children on a school trip, and the two pilots of the 757. The German Federal Bureau of Aircraft Accident Investigation (BFU) documented the sequence in report AX001-1-2/02, published in May 2004. Both aircraft were at FL360 on converging tracks in airspace controlled by Skyguide from Zurich. One controller was working two positions during a quiet night shift, an unofficial but tolerated practice. One radar system had been put into a degraded fallback mode for scheduled maintenance, which disabled the optical short-term conflict alert. The direct telephone lines were out for the same maintenance, so the neighbouring centre at Karlsruhe, whose own alert had triggered, could not get through to warn Zurich. Less than a minute before impact the controller instructed the Tupolev to descend. Seconds later both aircraft received resolution advisories. **The advisories were correct and correctly coordinated**: the Tupolev was told to climb, the 757 to descend. The Tupolev's crew followed the controller. The 757 followed its RA. Both aircraft descended, into each other. It is important to be exact about the failure mode, because it is easy to describe wrongly. Überlingen was **not** a coordination failure. The Mode S link did its job; the two senses were complementary. It was a **compliance** failure, and it demonstrates the same underlying truth from the other side: the protection is a two-aircraft protocol, and it only holds if both ends honour it. Whether the break is in the datalink or in the cockpit, the result looks the same — two aeroplanes manoeuvring the same way. The BFU identified as causal, alongside the organisational and technical failures at the air navigation service provider, that the regulations governing ACAS were **incompletely and partly contradictorily worded**. A near-collision over Suruga Bay, Japan, on 31 January 2001 — in which a Japan Airlines Boeing 747 was told by a controller to descend while its RA commanded a climb, the crew followed the controller, and about a hundred people were injured in the resulting violent avoidance manoeuvre — had already exposed the same ambiguity without producing an unambiguous rule. Überlingen produced one. Two things changed. First, the procedural rule quoted above was made explicit and mandatory worldwide. Second, the logic itself was amended, producing **version 7.1**, via two change proposals: - **CP112E — reversal logic.** Version 7.0 could fail to reverse an advisory in "vertical chase" geometries, where one aircraft manoeuvres opposite to its RA and the two end up travelling in the same vertical direction. Version 7.1 detects that the intruder is not doing what the coordinated solution assumed, and turns the advisory round: *"Climb, climb NOW"* or *"Descend, descend NOW"*. - **CP115 — the weakening annunciation.** Version 7.0's *"Adjust vertical speed, adjust"* was widely misread; monitoring found that in a substantial minority of cases crews increased their vertical rate when the system was asking them to reduce it. Version 7.1 replaces it with *"Level off, level off"*, which cannot be misunderstood. Version 7.1 was mandated in European airspace by Commission Regulation (EU) No 1332/2011, with retrofit required by 1 December 2015; ICAO Annex 10 required it for new installations from 1 January 2014 and for all ACAS II units from 1 January 2017. The microsim exposes both changes. Set coordination ON, issue the contradictory ATC instruction, tick the crew-follows-ATC box, and compare the two logic versions. With **version 7.1** the ownship detects the vertical chase and reverses, and in the default geometry the encounter still ends with well over a thousand feet of vertical miss. With **version 7.0** no reversal is issued and the same encounter ends at a few hundred feet — inside ALIM, and inside the margin the system exists to guarantee. Whether version 7.1 would have prevented Überlingen specifically is a modelling question rather than a settled fact, and reasonable analyses differ on how much time the reversal would have had; what is not in doubt is that the geometry which defeated version 7.0 is now handled. ## Controls -> what each maps to | Control | Symbol | Range and units | What it does | | --- | --- | --- | --- | | Relative heading | *φ* | 0 – 180° | Angle between the two ground tracks. 0° is a co-track overtake, 90° a perpendicular crossing, 180° head-on. Ownship always tracks 000° at a fixed 450 kt (about Mach 0.78 at FL350 in ISA); the intruder's track is offset by *φ*. | | Closure speed | \|d*r*/d*t*\| | 60 – 1200 kt | The initial range rate. The intruder's ground speed is solved from \|**v**\|² = *V*<sub>i</sub>² + *V*<sub>o</sub>² − 2*V*<sub>i</sub>*V*<sub>o</sub> cos *φ* and clamped to a realistic 110–700 kt, so some angle/closure pairs are unreachable — a 90° crossing at 450 kt cannot close slower than 450 kt. The HUD reports the **achieved** closure and flags it when clamped. | | Initial vertical offset | Δ*z* | −1200 to +1200 ft | Intruder altitude minus ownship altitude at *t* = 0. At Δ*z* = 0 the sense choice is a genuine tie and coordination is doing all the work; at ±400 ft or more the non-crossing rule decides it unaided. | | Ownship altitude | — | 2500 – 43 000 ft | Selects the **sensitivity level** and with it every threshold in the table above. Watch τ<sub>TA</sub>, τ<sub>RA</sub>, DMOD, ZTHR and ALIM all change together as the slider moves through the bands. | | Intruders | *N* | 1 – 5 aircraft | Additional traffic, staggered by 10 s of closest-approach time and 0.18 NM of horizontal miss each, on alternating vertical offsets. All of them alert; ownship coordinates with the most urgent. | | Mode S coordination | — | ON / OFF / transponder-only | ON is both aircraft TCAS II with the link running — senses complementary by protocol. OFF leaves each unit to decide alone. Transponder-only models an intruder that replies to interrogation but carries no TCAS, so it never manoeuvres. | | Logic version | — | 7.1 / 7.0 | 7.1 issues sense reversals in the vertical-chase geometry and annunciates *"Level off, level off"*. 7.0 does neither. | | Contradictory ATC instruction | — | button, key **A** | Gives the primary intruder a controller's instruction pointing the opposite way to its RA. Both commands are then drawn: green arrow for the RA, magenta for ATC. | | Crew flies the ATC instruction | — | checkbox, key **U** | The Überlingen error. The intruder's RA continues to be computed and displayed, but the aeroplane flies the controller's instruction instead. Off by default, because off is what the regulations require. | | Time compression | — | 0 – 8 × real time | Simulated seconds per wall-clock second; 0 is paused. The default is 4×. TCAS works in seconds, so real time is genuinely the right speed to watch it at least once. | | Vertical exaggeration | *k* | 10 – 140 × | Drawing only; no readout is affected. At true scale 600 ft of ALIM is 0.099 NM inside a picture ten miles wide, which is invisible. | | Seconds ruler | — | toggle | The tick marks on each intruder's ground track, one per second of remaining τ, coloured green above the TA threshold, amber between TA and RA, red inside the RA threshold. | | Range rings | — | toggle | Conventional 2 NM traffic-display rings around ownship. Deliberately present as a control experiment: a ruler in distance, alongside a ruler in time. | | *(readout)* τ to each intruder | *τ* | seconds | Range tau, −*r*/ṙ, per intruder, in the traffic table. | | *(readout)* Alert level | — | NONE / TA / RA | Highest advisory in force, with the TCAS display symbology drawn on the reference plane: open white diamond for other traffic, filled white for proximate (6 NM and 1200 ft), filled amber circle for a TA, filled red square for an RA. | | *(readout)* Projected vertical miss | — | ft, against ALIM | What the current advisory is expected to deliver at the closest point of approach, compared with the ALIM for this sensitivity level. This is the number that collapses when coordination is broken. | | *(readout)* Outcome at CPA | — | ft vertical and horizontal | The achieved miss once the pair has passed, flagged **NMAC** if under 500 ft horizontally and 100 ft vertically, and noting whether a reversal was issued. | ## Learning objective Come away able to say why a collision avoidance system reasons in seconds rather than miles, and to reproduce the argument from two numbers: three miles at 20 kt is nine minutes, eight miles at 1000 kt is under thirty seconds. Be able to write down modified tau and explain what DMOD is protecting against. Be able to state that an advisory requires a range test **and** an independent vertical test, and that the thresholds are not universal constants but a table indexed by altitude. Be able to explain sense selection as a projection against ALIM with a preference for the non-crossing sense — and then to explain why, at equal altitudes, that logic necessarily produces the same answer in both aircraft, and why the Mode S coordination link is therefore not an optimisation but the mechanism itself. Finally, be able to state the ATC conflict rule correctly and without hedging: the crew follows the resolution advisory, and following the controller instead is what turned Überlingen from a near miss into a collision. ## Limits and connections **What the model leaves out.** The primary encounter is built with a zero horizontal miss, which is the worst case rather than the typical one; real encounters usually have some lateral offset, and TCAS's own logic contains horizontal miss-distance filtering that is not reproduced here. Ground speeds are constant, there is no wind, and nothing turns — defensible, since TCAS II advisories are vertical only, but it means the geometry cannot evolve the way a real one does. Surveillance is idealised: there are no dropouts, no garbling, no interference limiting, and no bearing error at all, when in reality bearing error is the reason the whole system is vertical. **The TCAS logic here is a faithful skeleton, not a reimplementation.** The threshold table, modified tau, the vertical test, the ALIM projection, the coordination protocol and the DO-185B pilot response model are the real ones. The sense selection is a simplified version of the full logic, which contains additional machinery for altitude-crossing encounters, RA inhibits driven by aircraft performance, and multi-threat composite advisories. Multi-aircraft encounters here resolve only the ownship-to-intruder pairs; intruder-to-intruder coordination is not modelled, and real clusters of three or more TCAS aircraft compose pairwise in ways this does not capture. Sensitivity levels 2 and 3 are selected on **radio** altitude and are not modelled at all, which is why the altitude control starts at 2500 ft; the low-altitude RA inhibits are described above but not simulated. **The version 7.0 reversal deficiency is deliberately caricatured.** The model simply refuses to reverse when the intruder is moving in the same vertical sense as own aircraft. The real version 7.0 logic fails in a narrower and more specific set of geometries than that. The caricature is honest about direction — that geometry is the one CP112E was written to fix — but it should not be read as a description of the version 7.0 algorithm. **The pilot response is idealised.** Every crew in the simulation responds exactly on the DO-185B schedule. Real compliance is imperfect and known to be so: European monitoring programmes have consistently found a significant fraction of RAs flown late, weakly, or in the wrong sense, and improving compliance remains a larger lever on residual collision risk than any further change to the logic. The simulation's one non-compliant crew is the one you switch on deliberately. **Where the field is genuinely unsettled.** Whether version 7.1 would have prevented the Überlingen collision, given how late the encounter developed, is a modelling question on which analyses differ; the defensible claim is the narrower one, that the geometry which defeated version 7.0 is now handled. The transition to **ACAS X**, which replaces threshold logic with a policy computed offline by dynamic programming over a probabilistic encounter model, is expected to reduce both collision risk and nuisance alerts, but it is a different kind of system — its behaviour is a table rather than a rule, which makes it harder to explain and to certify, and the operational experience base is still thin. How collision avoidance should work for uncrewed aircraft, which cannot be assumed to have a pilot who responds in five seconds at a quarter of a g, is an active and unresolved area; ACAS Xu exists but the integration questions are not settled. **Connections.** TCAS sits on top of the [[Transponder]] and shares its interrogation frequencies with secondary surveillance [[Radar]]; it is the airborne complement to the ground service described in [[Air_traffic_control]]; it yields priority to the [[Terrain_awareness_and_warning_system]], because a warning about the ground outranks a warning about an aeroplane; on many modern types an RA can be flown by the [[Autopilot]] through an auto-flight mode that tracks the commanded vertical speed band; and it is one instance of the general design problem treated in [[Avionics]], of building a system that must be believed instantly and must almost never be wrong. ## References - RTCA, Inc. *DO-185B, Minimum Operational Performance Standards for Traffic Alert and Collision Avoidance System II (TCAS II), Version 7.1*. Washington, DC: RTCA, 2008. The defining document: threshold tables, modified tau, sense selection, coordination and the pseudocode of the collision avoidance logic. - RTCA, Inc. *DO-385, Minimum Operational Performance Standards for Airborne Collision Avoidance System X (ACAS Xa and ACAS Xo)*. Washington, DC: RTCA, 2018. - International Civil Aviation Organization. *Annex 10 to the Convention on International Civil Aviation: Aeronautical Telecommunications, Volume IV — Surveillance and Collision Avoidance Systems*. Montreal: ICAO. ACAS II standards and recommended practices, including the version 7.1 applicability dates. - International Civil Aviation Organization. *Annex 6, Operation of Aircraft, Part I — International Commercial Air Transport: Aeroplanes*. Montreal: ICAO. ACAS II carriage requirement by mass and passenger capacity. - International Civil Aviation Organization. *Doc 8168, Procedures for Air Navigation Services — Aircraft Operations (PANS-OPS), Volume I: Flight Procedures*. Montreal: ICAO. ACAS II operating procedures, including the requirement to follow an RA in preference to a conflicting ATC instruction. - International Civil Aviation Organization. *Doc 4444, Procedures for Air Navigation Services — Air Traffic Management (PANS-ATM)*, 16th edition. Montreal: ICAO, 2016. Chapter 15, ACAS procedures for controllers, including the transfer of separation responsibility once an RA is being flown. - Eurocontrol. *ACAS II Guide: Airborne Collision Avoidance System II (incorporating TCAS II version 7.1)*. Brussels: Eurocontrol, current edition. The standard operational reference, and the source usually quoted for the sensitivity-level table. - Eurocontrol. *ACAS II Bulletin* series. Brussels: Eurocontrol. Short operational notes on RA compliance, reversal logic and version 7.1 behaviour, drawn from European monitoring programmes. - Federal Aviation Administration. *Introduction to TCAS II Version 7.1*. Washington, DC: FAA, February 2011. The concise official description of the version 7.1 changes, CP112E and CP115. - Bundesstelle für Flugunfalluntersuchung (German Federal Bureau of Aircraft Accident Investigation). *Investigation Report AX001-1-2/02*: mid-air collision near Überlingen, Lake Constance, 1 July 2002. Braunschweig: BFU, May 2004. - Aircraft and Railway Accidents Investigation Commission (Japan). *Aircraft Accident Investigation Report: Japan Airlines Flight 907 and Japan Airlines Flight 958, near Yaizu, Shizuoka Prefecture, 31 January 2001*. Tokyo: ARAIC, 2002. The Suruga Bay near-collision, in which a crew followed a controller against an RA. - European Union. *Commission Regulation (EU) No 1332/2011 of 16 December 2011 laying down common airspace usage requirements and operating procedures for airborne collision avoidance*. Official Journal of the European Union, L 336, 20 December 2011, as amended by Regulation (EU) 2016/583. - United States. *Code of Federal Regulations, Title 14, Part 91*, §91.123 (compliance with ATC clearances and instructions, including the TCAS resolution advisory exception) and Part 121, §121.356 (TCAS II equipage for air carriers). - Kuchar, James K., and Ann C. Drumm. "The Traffic Alert and Collision Avoidance System." *Lincoln Laboratory Journal*, vol. 16, no. 2, 2007, pp. 277–296. The clearest published engineering account of the alerting logic and its design trade-offs. - Kochenderfer, Mykel J., Jessica E. Holland, and James P. Chryssanthacopoulos. "Next-Generation Airborne Collision Avoidance System." *Lincoln Laboratory Journal*, vol. 19, no. 1, 2012, pp. 17–33. The ACAS X formulation as a partially observable Markov decision process. - Holland, Jessica E., Mykel J. Kochenderfer, and Wesley A. Olson. "Optimizing the Next Generation Collision Avoidance System for Safe, Suitable, and Acceptable Operational Performance." *Air Traffic Control Quarterly*, vol. 21, no. 3, 2013, pp. 275–297. - Livadas, Carolos, John Lygeros, and Nancy A. Lynch. "High-Level Modeling and Analysis of the Traffic Alert and Collision Avoidance System (TCAS)." *Proceedings of the IEEE*, vol. 88, no. 7, July 2000, pp. 926–948. A formal-methods treatment of the coordination protocol, including the address tie-break. - Williamson, Tom, and Neil A. Spencer. "Development and Operation of the Traffic Alert and Collision Avoidance System (TCAS)." *Proceedings of the IEEE*, vol. 77, no. 11, November 1989, pp. 1735–1744. Contemporary account of the design decisions, including why horizontal advisories were not pursued. **On the spine:** [[Avionics]] · [[Fly-by-wire]] · [[Autopilot]] · [[Head-up_display]] · [[Traffic_collision_avoidance_system]] · [[Terrain_awareness_and_warning_system]] · [[Radio_navigation]] · [[Radiation_hardening]] · [[Radar]] · [[Air_traffic_control]] · [[Aviation]].  **Part of the [[Avionics]] hub** — main article for section X18, *Collision-avoidance systems*. Related sections: [[Terrain_awareness_and_warning_system]] · [[Air_traffic_control]].  ## Wikipedia : Wikitube **Strict pair:** [Wikipedia](https://en.wikipedia.org/wiki/Traffic_collision_avoidance_system) : [Wikitube](https://en.wikitube.io/wiki/Traffic_collision_avoidance_system) --- *PORTAL_Avionics three.js batch · 2026-08-05 · sim staged in `_3d_deploy_stage/`.*