ACP WGN WP 16 February 2016 AERONAUTICAL COMMUNICATIONS PANEL (ACP) WORKING GROUP N (NETWORKING) Montreal, Canada, 11-20 May, 2005 CPDLC Message Sequencing Issues Presented by EUROCONTROL Summary At the New Orleans ACP/WGN meeting (November 2004), a last minute modification was made to the PM-CPDLC proposal to include a message sequence number in the scope of the PM-CPDLC Application Checksum. Since then, the Eurocontrol Link2000+ Team has conducted an extensive review of the need for such a sequence number both from the communications and operations perspectives. This paper records the result of this review and concludes with a strong recommendation that the proposed message sequence number is removed from PM-CPDLC. 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 i TRS164/TW/3/3/43 CONTENTS Page 1 WHAT IS MEANT BY “MESSAGE SEQUENCING”? .......................................................................... 1 2 SAFETY CONSIDERATIONS .................................................................................................................. 6 3 CONCLUSION AND RECOMMENDATIONS ....................................................................................... 8 3.1 SUMMARY OF EUROCONTROL LINK2000+ REVIEW.................................................................................... 8 3.2 RECOMMENDATIONS TO WGN ................................................................................................................... 9 A OPERATIONAL ANALYSIS OF OUT-OF-SEQUENCE MESSAGES .............................................. 11 16/02/2016 Draft 0.4 ii What is meant by “Message Sequencing”? 1 What is meant by “Message Sequencing”? There has been a general concern that the delivery, by CPDLC, of “out-of-sequence messages” could lead to an operational hazard. The first task of the Eurocontrol review was thus to investigate whether this was true and if so, what is the underlying hazard. When considering this, a useful input paper was found to be a paper submitted, by Boeing, to the Industry Rapid Reaction Task Force (IRRF), which was responsible for preparing a report to Link2000+ on FANS-1/A accommodation. This paper is attached as Annex A, and is particularly useful because it contains many examples of operational scenarios where CPDLC and ADS messages are delivered out-of-sequence, and the operational impact of these scenarios What is clear from these examples is that not only are there many cases where out-ofsequence delivery does not lead to a hazard, but in-sequence delivery is not itself sufficient to avoid a hazard. A hazard results only if the pilot executes dependent clearances in a different order to that intended by the controller, In fact, what is operationally significant is whether the dependency and the order of execution intended by the sending controller is the same as that perceived by the receiving aircrew. This view was backed up by controllers from various regions, with the emphasis placed on the “order of execution” rather than the “order of transfer”. The operational requirement can therefore be summarised as: 1. The controller’s intent as to the order of execution of dependent clearances is clearly and unambiguously communicated to the pilot. There are of course related requirements in that the pilot is required to execute clearances in the indicated order, and on the controller to give clear, precise and unambiguous instructions. However, this is no different to the situation with voice based controller-pilot communications. The IRRF paper and Maastricht controllers also argued against an over-prescriptive approach to message sequencing. In particular: ● Aborting a CPDLC connection can itself lead to a hazard due to the impact on controller and aircrew workload, and confusion over which clearances have and have not been delivered. Hence, aborting a CPDLC connection due to out-ofsequence message delivery when there was no dependency between the messages delivered out-of-sequence could induce an unnecessary hazard. ● Delaying delivery of one message in order to ensure in-sequence delivery of an earlier message that had to be retransmitted is often highly undesirable when there is no dependency between the messages. Discarding such a message without presentation to the intended recipient can be equally or even more undesireable. This is because the unnecessary delay can have an operational impact. For example, a message that tells you a wheel has fallen off is equally valid whether it arrives in sequence or out of sequence and discarding it would be operationally dis-advantageous. From this, it is clear that the mechanism for communicating the correct order of execution of dependent clearances should not itself impede the exchange of independent CPDLC messages. It should be noted that many CPDLC messages are independent of each other 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 1 TRS164/TW/3/3/43 and can hence be communicated in different relative orders without significantly affecting the outcome. There are thus further operational requirements: 2. The mechanism for communicating the correct order of execution of dependent clearances should not itself impede the exchange of independent CPDLC messages. and 3. Operationally meaningful data shall not be discarded without human review or delayed on the sole basis of its delivery being out of the order of transmission. Requirement 2 is on the mechanism used to convey execution order, while requirement 3 deals with messages that arrive out of order. In general, discarding or delaying messages just because they arrive out of transmission order is undesirable because this can delay operationally valid information. Having established the requirements, it was then necessary to consider how the requirements are satisfied. It is understood that, during the development of CPDLC, three possible strategies were discussed for communicating dependent clearances. Strategy #1: One Clearance at a time Under this strategy, which is common with voice communications, and with all current operational CPDLC implantations (ATN and FANS-1/A), dependent clearances are issued by a controller, one at a time and in the intended order of execution. The controller waits for a closure response before issuing the next clearance in the sequence. However, an independent CPDLC transaction may be opened at any time during the transactions used to transfer the sequence of dependent clearances. This strategy ensures the order of execution of dependent clearances, as the controller waits for the pilot to confirm execution of each clearance before issuing the next one. If the pilot responds UNABLE to any clearance in the sequence, the controller is responsible for determining the course of action resulting from a partially completed sequence. Requirement 1 is satisfied by waiting for a closure response before the next clearance is issued. Requirement 2 is satisfied because the controller is allowed to start an independent transaction during the sequence of dependent transactions. The procedure only applies to CPDLC transactions of the same type. Requirement 3 is also satisfied as there is no need to discard or delay an out of order transmission; the communications service is not relied upon to maintain transmission order1. Strategy #2: Concatenated Messages Under this strategy, the controller issues a sequence of dependent clearances as a single concatenated CPDLC message. There are two cases to consider: 1 However, use of the connection mode transport protocol for ATN CPDLC means that an out of order message will be delayed in the network. 16/02/2016 Draft 0.4 2 What is meant by “Message Sequencing”? a) Where the clearances are dependent, in the sense that the pilot must give a single response to all of them, but the execution order is not important, and b) Where the clearances are dependent, a single response must be given to all of them, and the execution order is important. An example of the first case is a controller concatenating “Cleared direct to <fixname>, Climb to Flight level 350” into a single CPDLC message. These are dependent clearances, and the controller wants the crew to do both or neither of them. However, the execution order is not significant, They are most likely to be performed in parallel. Use of the second case is more controversial as there is concern that CPDLC semantics are not well enough understood to permit it. In Europe, the LINK2000+ Operational Focus Group has, for such reasons, issued initial guidance that these type of clearances are not to be used. Instead, when there is a required execution order, the controller should always use Strategy #1 above. The controversy is over the semantic of message element UM165 THEN and whether it unambiguously implies an order of execution. For example, a controller may concatenate “Climb to FL370, Then, Turn Left 30°”. In this case, the concatenation is again used to signal the dependency but, additionally, the message element UM165 THEN is used to indicate that there is a required order of execution, and that the order of execution is the same as the order of the message elements within the concatenated CPDLC message. This approach uses concatenation and UM165 THEN to convey the controller’s intent as to both the dependency and the order of execution of dependent clearances, and relies upon this interpretation of UM165. As such, the approach meets requirement 1. CPDLC requires that a single closure response is given and hence the pilot must either accept and execute the sequence of clearances in its entirety or reply UNABLE. In either case, the controller is given a clear indication as to the outcome of the sequence of clearances and can decide on the next action. Independent CPDLC transactions may be opened before the closure response is received and hence requirement 2 is also satisfied. Requirement 3 is also satisfied as there is no need to discard or delay an out of order transmission; the communications service is not relied upon to maintain transmission order1. If more dependent clearances need to be issued than can be concatenated into a single CPDLC message, then this strategy can be combined with strategy #1. This approach appears workable, however, the use of concatenation for dependent clearances where the order of execution is significant, awaits agreement on the semantics of UM165 and is thus considered as currently “for further study”. Strategy #3: Transmission Order Under this strategy, dependent clearances are sent as separate CPDLC messages, rather than being concatenated into the same message. The actual transmission order of the separate CPDLC messages is used to imply the sequence of execution, with the controller no longer waiting for a closure response for a clearance before issuing a subsequent dependent clearance. This is significantly different from the previous strategies where dependency and execution order is given by the semantics of the CPDLC messages themselves. Under this strategy, the communications service is being used and relied upon to indicate dependency and execution order. This has important safety implications, 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 3 TRS164/TW/3/3/43 demanding a much higher level of assurance from the Communications Service than is required by the other strategies. However, there are a number of issues associated with such a strategy: 1. How is the dependency implied given that by no means all CPDLC messages are dependent on each other? How does the pilot know that two successive but separately sent messages have a required execution order? Use of UM165 THEN does not appear appropriate here. For example, how would a flight crew react to PROCEED DIRECT TO <fixname> + THEN? They could not WILCO the message as the clearance is incomplete. UM165 THEN cannot correctly be used on its own. It could be used as in, for example, THEN + PROCEED DIRECT TO <fixname>, but this is not really compatible with voice phraseology which would naturally be more explicit (e.g. AT FL350 PROCEED DIRECT TO <fixname>). This is a CPDLC message (UM75), but in this case the dependency is being signalled by the message itself and this is a different strategy to relying the Communications Service. 2. What happens if a pilot responds UNABLE to a message before a subsequent and dependent clearance arrives? Does the pilot have to respond UNABLE to that message as well, and how does the pilot know this? It would seem that the pilot has to assume that each instruction has to be executed in sequence and if an UNABLE response is given then all subsequent instructions must be responded to with UNABLE until the controller advises otherwise (although it is not clear as to how this advice can be given with the current CPDLC message set), and this is, anyway, unlikely to result in operational satisfaction for either aircrew or controller. 3. ICAO Doc9705 2.3.7.3.2.2 states “When a CPDLC-user queues received messages, messages with the highest Urgency type shall be placed at the beginning of the queue”. Thus, mixing dependent clearances with different Urgency attributes can result in a hazard as an ICAO-compliant system may require that controller’s intended order of sequence is incorrectly displayed to the pilot. The controller must hence ensure that messages with different Urgency attributes are always independent of each other. The underlying problem with this strategy is that CPDLC lacks a mechanism to group separate messages together and only the controller is really aware of any dependency between successive messages. The pilot thus has to assume that all messages are dependent on the preceding message. Requirement 1 is met but only by ensuring that requirements 2 and 3 cannot be met. This is because the receiver must always assume that successive messages are dependent on each other and must be executed in reception order. There is thus no scope for mixing independent and dependent instructions as responding UNABLE to one message means that UNABLE must be the response to subsequent messages, even when they are independent. It may be possible for the pilot to exercise a degree of judgement on whether a received message is dependent on the outcome of a previous message but this is potentially hazardous as, unless the relationships between every possible message are clearly defined, there is always the risk of mis-understanding. On the other hand, the HMI cannot permit out of order display to the pilot as, by definition, the preceding message has not been received and it is not possible to determine whether there was a dependency on that message. 16/02/2016 Draft 0.4 4 What is meant by “Message Sequencing”? This strategy does not meet the requirements identified above. The lack of any apparent mechanism to signal the end of a dependent sequence may also make it unworkable in practice. Conclusion The first two strategies are sufficient for CPDLC, and meet the known operational requirements for message sequencing in both congested and uncongested airspace. The third strategy does not meet requirements 2 and 3, and only meets requirement 1 by imposing onerous procedures on both controller and pilot. Any deviation from these procedures could quickly result in a hazard. Strategy #3 is the only one to require that the communications service preserves the message sequence. Indeed that is its problem. The issues are caused by not using CPDLC message element semantics to identify dependencies and required execution order. It is believed that strategies #1 and #2 are sufficient for world-wide use of CPDLC. CPDLC is not mature enough to support Strategy #3 and any further consideration of Strategy #3 should be for further study. There may be scope for developing a strategy for issuing dependent clearances in separate messages, using messages such as UM75 to signal the dependency. This moves the semantics of the dependency back into CPDLC and should not rely on the Communications Service to indicate dependency. The conclusion is that there is no workable CPDLC strategy that requires the Communications Service to maintain message sequence. Dependency semantics can and always should be part of CPDLC itself and should not rely on features of the Communications Service. 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 5 TRS164/TW/3/3/43 2 Safety Considerations Eurocae ED-120/RTCA DO-290 “Safety and Performance Requirements Standard for Air Traffic Data Link Services in Continental Airspace (Continental SPR Standard)” contains the Safety Objectives and Requirements for continental deployment of en-route CPDLC. For the ATC Clearance Service (ACL), the Safety Objective relevant to this discussion is SO-ACL-15 “The likelihood of undetected out of sequence messages used for separation shall be no greater than remote.” Following the above analysis, and consideration of requirement 1 above, SO-ACL-15 should be interpreted as “the likelihood of out of order execution of a sequence of dependent clearances shall be no greater than remote”. This “remote” Safety Objective is important for deployment as implementers will have to demonstrate to a Safety Authority that the probability that this will occur is remote. Both the design of CPDLC and the software architecture will have to be taken into account when approval for operational use is granted. Eurocae ED-120/RTCA DO-290 also contains a number of Safety Requirements. Those that are linked to SO-ACL-15 are: SR-ACL-1 When a clearance requires execution of more than one maneuver to be done in a specific sequence, the clearances shall be put in the order that they are to be executed in a single uplink message. SR-ACL-2 Each uplink message shall be uniquely identified for a given aircraftATSU pair. SR-ACL-3 Each downlink message shall be uniquely identified for a given aircraftATSU pair. SR-ACL-4 A response message shall indicate to which message it refers. SR-ACL-7 Messages shall be responded to in their entirety. SR-ACL-8 Any processing (data entry/encoding/transmitting decoding/displaying) shall not affect the intent of the message. SR-ACL-22 Messages shall be transmitted/received in the order that they are sent. SR-ACL-23 Only one dialogue of a given type shall be open at any time for a given aircraft-ATSU pair. SR-ACL-2, SR-ACL-3, SR-ACL-4 and SR-ACL-8 are common requirements for many safety objectives and are clearly necessary for correct interpretation of CPDLC messages. CPDLC itself complies with these requirements and implementers are required to demonstrate that their software implements these features correctly. SR-ACL-23 is satisfied by the implementation of strategy #1 above. Likewise, SR-ACL-1 and SR-ACL-7 are satisfied by the implementation of strategy #2. 16/02/2016 Draft 0.4 6 Safety Considerations SR-ACL-22 appears to be specific to strategy #3 and is a necessary condition for the implementation of strategy #3, but only if the possible CPDLC defect identified in the previous section is ignored. These three groups of safety requirements appear to thus align with the strategies presented earlier and, upon analysis, appear to be alternative and possibly complementary ways of meeting the Safety Objective, rather than all needing to be satisfied simultaneously. In particular, given the analysis presented in the preceding section, SR-ACL-22 is probably inappropriate given that the corresponding strategy is not believed to be workable and a change to CPDLC in order to resolve this will result in a very different safety requirement. This analysis has been presented to representatives of Eurocae WG-53/RTCA SC-189 and has been accepted. A change proposal to ED-120 is thus anticipated in order to bring it into line with the analysis presented in this paper. In summary, CPDLC with either strategy #1 or strategies #1 and #2 meets the ED120/RTCA DO-290 Safety Objectives and are sufficient to ensure that dependent clearances can be communicated using CPDLC. The Safety Requirement that appears to require that the communications service maintains message sequence only applies to strategy #3 and is only there because of a defect in CPDLC. It is therefore an inappropriate and unnecessary requirement. 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 7 TRS164/TW/3/3/43 3 Conclusion and Recommendations 3.1 Summary of Eurocontrol Link2000+ Review The following is the agreed conclusion of the Link2000+ Message Sequencing discussion, as documented and agreed by Link2000+: 16/02/2016 The “message out of sequence” hazard occurs when a sequence of dependent clearances is executed out of sequence i.e. the pilot executes the clearances in a different sequence to that intended by the controller. This is the agreed interpretation of ED-120/DO-290 H-ACL-15 . The Safety Objective is therefore to ensure that the controller’s intent as to the order of execution of dependent clearances is preserved during the execution of the clearances by the pilot. The above is viewed as the correct interpretation of SO-ACL-15 (The likelihood of undetected out of sequence messages used for separation shall be no greater than remote). When a controller needs to issue a sequence of dependent clearances, the controller is required to issue that clearance as a single concatenated CPDLC message; the order in which the clearances occur in the message specifies the order of execution. This requirement is echoed by ED-120 SR-ACL-1 (When a clearance requires execution of more than one manoeuvre to be done in a specific sequence, the clearances shall be put in the order that they are to be executed in a single uplink message.) ED-120 SR-ACL-7 (Messages shall be responded to in their entirety) ensures that the pilot must provide a single unambiguous response to a sequence of dependent clearances. When a controller needs to issue a clearance that is dependent on the successful outcome of a clearance that had been issued earlier, they are required to wait until the execution of that earlier clearance has been completed before the new clearance is issued. This outcome may be determined by a pilot report or through surveillance information. ED-120 SR-ACL-23 (Only one dialogue of a given type shall be open at any time for a given aircraft-ATSU pair) avoids potential confusion that could result from a controller issuing a new clearance before the pilot had responded to an earlier clearance. . ED-120 SR-ACL-23 does not prevent a controller issuing a different type of instruction (e.g. an SSR Code Change) before the response to a clearance has been received. The response to (e.g. an SSR Code Change) may also be received before the response to the clearance. Neither the communications system nor the HMI should prevent this behaviour i.e. out of sequence responses to CPDLC messages of different dialogue types must not be prevented. It is noted that SR-ACL-2 (Each uplink message shall be uniquely identified for a given aircraft-ATSU pair), SR-ACL-3 (Each downlink message shall be uniquely identified for a given aircraft-ATSU pair), SR-ACL-4 (A response message shall indicate to which message it refers), SR-ACL-8 (Any processing (data entry/encoding/ transmitting/ decoding/displaying) shall not affect the intent of the message) are also required in order to meet SO-ACL-15. However, they are otherwise outside of the scope of this discussion. Draft 0.4 8 Conclusion and Recommendations It is noted that ICAO Doc9705 2.3.7.3.2.2 (When a CPDLC-user queues received messages, messages with the highest Urgency type shall be placed at the beginning of the queue) permits the receiver to reorder messages by Urgency and that this does not have an interoperability or safety impact. It is noted that there are also situations where a it is desirable that a CPDLC Message is uplinked before the response to an earlier message has been received. For example, where a change in SSR Code is required shortly after a clearance has been issued. The meaning of “dialogue type” in ED-120 needs to be clarified to ensure such cases are fully understood. It is noted that the OFG has recommended that when dependent clearances are concatenated into the same CPDLC message, they should be separated by the message element UM165 THEN. It is thus concluded that: (a) ensuring the execution order of dependent clearances by the pilot is the same as that intended by the controller, is a Safety Requirement, (b) preserving the transmission order of CPDLC messages themselves, on receipt, is not a Safety Requirement. This is because H-ACL-15 is fully mitigated by SR-ACL-1 and SR-ACL-7 (together with SRACL-2, 3, 4 and 8) for clearances that require execution of more than one manoeuvre in a specific sequence. SR-ACL-23 also avoids potential confusion that could result from a controller erroneously issuing a clearance of the same dialogue type of an earlier CPDLC message and for which a response was outstanding. ED-120 SR-ACL-22 (Messages shall be transmitted/received in the order that they are sent) thus appears to have no useful role to play. It is proposed that SR-ACL-22 should no longer reference SO-ACL-15 and that the reference from SR-ACL-22 to SO-ACL-15 should be deleted. 3.2 Recommendations to WGN ICAO ACP/WGN is invited to: 1. Review the analysis presented in this paper. 2. Remove the addition of a message sequence number from the scope of the PMCPDLC checksum, as there is no requirement on the communications service to maintain message sequence. The recommendation to remove the message sequence number from the scope of the PMCPDLC checksum is a strong recommendation, not only because it is unnecessary but also because its implementation will adversely affect ground CPDLC implementations that place the ATN End System in a separate front end processor or server (which is the case with all known implementations). This is because the PM-CPDLC checksum should be generated when the message is first originated if it is to give the maximum safety benefit, and then communicated end-to-end without modification. However, in a distributed architecture there are potentially several points of message origination and messages may be originated by both automatic and manual processes. A shared sequence number is difficult to manage in such a system and regeneration of the 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 9 TRS164/TW/3/3/43 checksum in the front end processor is the only realistic implementation strategy, thus adding an additional point at which a failure could occur. 16/02/2016 Draft 0.4 10 Conclusion and Recommendations A Operational Analysis of Out-Of-Sequence Messages Acknowledgement: This annex contains an updated version of a paper with the above title prepared by Boeing ATM and submitted to the IRTF for Link2000+. A.1 Introduction One of the concerns over use of FANS-1/A (reference: ED100/DO258) in domestic airspace is the lack of a systematic prevention of out-of-sequence datalink message delivery. This is not currently viewed as an issue in oceanic environments using FANS-1/A, so Boeing wanted to understand the operational requirement better in order to arrive at a solution that would satisfy the domestic environment without jeopardizing oceanic FANS-1/A utility. To understand the requirement better, operational analysis was conducted, as outlined in this paper. Operational scenarios involving out-of-sequence message delivery were laid out, and then two aspects were considered for each scenario: 1. We listed the operational impact of allowing the out-of-sequence deliveries to occur. 2. We listed the operational effects of introducing a communications layer solution for assuring in-sequence delivery to upper layers of automation (i.e. sequence of delivery is assured across the communications ‘pipe’, but is not assured or within any automation or systems above the ‘pipe’, to include the CPDLC and ADS applications themselves). With this solution, the solution was designed to deliver messages in sequence, but should any messages be received out of sequence the messages would be discarded without notification to the recipient and the communications ‘pipe’ removed without further use. Note: This particular solution to in-sequence delivery assurance was specifically examined because it is a solution that has been proposed for resolving this issue, and its impact on oceanic operations needed to be looked at to ensure it would be acceptable in existing datalink operational regions. To completely understand the context, we started with a voice-only scenario. This may seem strange for a datalink analysis, but it helped understand more clearly both the operational hazard and the potential solutions. A.2 Operational Context The following scenarios concern some out-of-sequence exchanges that could arise from using CPDLC and ADS. The operational context for these sequences is: 1. Congested, continental en-route airspace. 2. Mode C radar surveillance is available. 3. VHF R/T channel is available, and is the primary communications tool for tactical communications, communications requiring immediate responses, and communications involving complex constructs and instruction sequences. 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 11 TRS164/TW/3/3/43 Note: The term “complex constructs and instruction sequences” in the above context is not intended to include route clearances and similar lengthy messages. Such exchanges are complex in the sense that a lot of data is exchanged, which lend themselves very well to data link. The above refers to complexities in the sense that subtle operational nuances, dependencies, and meanings are being conveyed in a format and sequence not routinely seen on the ground or in the air. 4. Primary use of data link is CPDLC for routine communications only. 5. CPDLC is used strategically. Data link is not to be used for immediate separation purposes under normal circumstances. 6. No voice readback is required for any message. A.3 Out-Of-Sequence Message Scenarios A.3.1 Dependant Clearances, Delivered in Sequence via Voice In the following sequence, two clearances are sent in rapid sequence via voice, with the second clearance dependant upon the first. Using voice the clearances arrive in sequence, but are operationally acknowledged out-ofsequence. A.3.1.1 Sequence No. 1. Via R/T: [point]” ATSU Aircraft “cleared direct to 2. Via R/T: “climb to and maintain [level]” A.3.1.2 Comment This clearance is only to be executed if the aircrew accepts the direct routing. 3. Via R/T: “Wilco, climb to [level].” 4. Via R/T: “Unable, direct to [point].” Conclusion The above “voice only” scenario was included to clarify the operational situation being discussed. Two points are apparent: 1. Even when messages arrive in sequence (voice or data link), unless a stipulation is placed on the crew to answer them in sequence, the out-of-sequence hazard still exists. This, in turn, means that simply embedding an “in-sequence-delivery” function in the communications chain does not resolve the envisaged hazard. 16/02/2016 Draft 0.4 12 Conclusion and Recommendations 2. The operational need and desirability of the above scenario should be confirmed during the hazard analysis for out-of-sequence delivery of messages. A common reaction to the above scenario during draft internal and external reviews was to question its operational viability. Controllers would normally either wait for the acknowledgement of the first clearance to issue a second, dependent clearance, or they would choose a clearer construct (e.g. “Cleared direct to [position], when steady on track, climb to [level]”). Both of these reactions would appear to apply to data link as well, with the latter equating to a ‘concatenated’ CPLDC uplink. The overall conclusion is that the operational viability of this “out-of-sequence” scenario needs to be clarified. If it proves viable, then a communications-level in-sequence assurance mechanism is insufficient to resolve the scenario. A.3.2 Dependant Clearances, Delivered in Sequence via Data link In the following sequence, two clearances are sent in rapid sequence via data link, with the second clearance dependent upon the first. As with the preceding scenario, the clearances arrive in sequence, but are operationally acknowledged out-of-sequence. A.3.2.1 Sequence No. A.3.2.2 ATSU Aircraft 1 Uplink “cleared [point]” direct 2 Uplink “climb to and maintain [level]” Comment to This clearance is only to be executed if the aircrew accepts the direct routing. 3 Wilco, climb to [level]. 4 Unable, direct to [point]. Conclusion This scenario demonstrates the same two points as the voice scenario (ref:0), this time in a data link context: 1. Even when messages arrive in sequence (voice or data link), unless a stipulation is placed on the crew to answer them in sequence, the out-of-sequence hazard still exists. This, in turn, means that simply embedding an “in-sequence-delivery” function in the communications chain does not resolve the envisaged hazard. 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 13 TRS164/TW/3/3/43 2. The operational need and desirability of the above scenario should be confirmed during the hazard analysis for out-of-sequence delivery of messages. Controllers should either wait for the acknowledgement of the first clearance to issue a second, dependent clearance, or they should choose a clearer construct (e.g. a single concatenated CPDLC message containing both clearances with the dependency clear in the message, e.g. “cleared direct to [point] then climb to and maintain [level]”). The overall conclusion is that the operational viability of this “out-of-sequence” scenario needs to be clarified. If it proves viable, then a communications-level in-sequence assurance mechanism is insufficient to resolve the scenario. A.3.3 Dependant Clearances, Delivered and Answered in Sequence via Data link In the following sequence, two clearances are sent in rapid sequence via data link, with the second clearance dependent upon the first. As with the preceding scenario, the clearances arrive in sequence. However, in this scenario the clearances are also operationally acknowledged in sequence, so all parts of the end-to-end chain handled the messages in exactly the sequence intended by the issuing controller. A.3.3.1 Sequence No. A.3.3.2 ATSU Aircraft 2 Uplink “climb to and maintain [level]” 2 Uplink “cleared [point]” direct Comment This clearance is only to be executed if the aircrew accepts the subsequent direct routing. to 3 Wilco, climb to [level]. 4 Unable, direct to [point]. Conclusion Like the first to scenarios, this scenario (ref: 0 and 0) demonstrates that the underlying hazard is not resolved by in-sequence message delivery. 1. Even when messages arrive in sequence (voice or data link), and the crew answers them in sequence, the underlying hazard still exists. This, in turn, means that simply embedding an “in-sequence-delivery” function in the communications chain does not resolve the envisaged hazard. 16/02/2016 Draft 0.4 14 Conclusion and Recommendations 2. The operational need and desirability of the above scenario should be confirmed during the hazard analysis for out-of-sequence delivery of messages. Controllers should either wait for the acknowledgement of the first clearance to issue a second, dependent clearance, or they should choose a clearer construct (e.g. a single concatenated CPDLC message containing both clearances with the dependency clear in the message, e.g. “cleared direct to [point] then climb to and maintain [level]”). The overall conclusion is that the operational viability of this scenario needs to be clarified. If it proves viable, then a communications-level in-sequence assurance mechanism is insufficient to resolve the hazard. A.3.4 Dependent Clearances, Delivered Out-Of-Sequence In the following sequence, two clearances are sent in rapid sequence via data link, with the second clearance dependent upon the first. The clearances arrive out-of-sequence, and are operationally acknowledged in their order of arrival. A.3.4.1 Sequence No. ATSU Aircraft Comment 1. Uplink “cleared direct to [point]” 2. Uplink “climb maintain [level]” A.3.4.2 to and This clearance is only to be executed if the aircrew accepts the direct routing. 3. Aircrew receives and Wilco’s climb to [level]. 4. Aircrew receives and Unable’s cleared direct to [point]. Conclusion The above scenario results in exactly the same operational hazard as the first two, though the cause is different. In this scenario the cause of the hazard is that the clearances were delivered out of sequence, while in the first two they were simply acknowledged out of sequence. 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 15 TRS164/TW/3/3/43 1. This scenario demonstrates that the hazard is in out-of-sequence display and acknowledgement of the uplinks. This hazard arrives solely through the use of dependent and unacknowledged clearances by the controller. Once the messages are identified as “out-of-sequence”, system-level functions could be introduced to take the appropriate operational action (e.g. display to the crew only in order). 2. The operational need and desirability of the above scenario should be confirmed during the hazard analysis for out-of-sequence delivery of messages. Controllers should either wait for the acknowledgement of the first clearance to issue a second, dependent, clearance, or they should choose a clearer construct (e.g. a single concatenated CPDLC message containing both clearances with the dependency clear in the message, e.g. “cleared direct to [point] then climb to and maintain [level]”). The overall conclusion is that the operational viability of the “out-of-sequence” scenario needs to be clarified. If it proves viable, then a communications-level in-sequence assurance mechanism is insufficient to resolve that scenario. A.3.5 Dependent Clearances with Conflicting Alerting Attributes In the following sequence, two clearances are sent in rapid sequence via data link, with the second clearance dependent upon the first. The clearances arrive in sequence, but because of the operationally defined alerting attributes in the FANS or ATN specifications, they are operationally acknowledged out-ofsequence. A.3.5.1 Sequence No. 16/02/2016 ATSU Aircraft 1 Uplink “cleared [point]” direct 2 Uplink “immediately climb to and maintain [level]” Comment to This clearance is only to be executed if the aircrew accepts the direct routing. 3 Aircrew Wilco’s “immediately climb to and maintain [level].” Aircrew reacts to the higher alerting, and responds to this message first. 4 Aircrew Unable’s “cleared direct to [point].” Message received prior to above, but answered second. Draft 0.4 16 Conclusion and Recommendations A.3.5.2 Conclusion The above scenario results in exactly the same operational hazard as the first two, though the cause is different. In this scenario the cause of the hazard is that the clearances were delivered in-sequence, but the message alerting attributes drew the attention of the crew to the second message and they responded to the two messages out-of-sequence. 1. This scenario demonstrates that the hazard is in display and out-of-sequence acknowledgement of the uplinks. This hazard arrives solely through the use of dependent and unacknowledged clearances by the controller. System-level functions could be introduced to take the appropriate operational action (e.g. prevent the crew from acknowledging the messages out of order). 2. The operational desirability of preventing the crew from answering messages out-ofsequence should be analyzed. For dependant clearance cases, controllers should wait for the acknowledgement of the first clearance to issue a second, dependent, clearance. 3. This scenario may indicate contradictory provisions should both alerting attributes insequence delivery be required. If all messages are expected to be dealt with in the order of issuance by the controller, than different alerting attributes for messages may be inappropriate, as they appear to be indicating that crews should respond to some messages out-of-sequence. Similar contradictions may be present in the use of urgency attributes for messages. If two messages are delivered in sequence, but then reversed during queuing and transit to aircrew HMI, as required by datalink specifications, the end-to-end effect is still outof-sequence delivery. Although communications layers delivered the messages in sequence, upper layers delivered them out-of-sequence as specified. The overall conclusion is that a communications layer solution will not resolve this scenario, and may in fact be overridden by other parts of the same datalink specification set which then contributes to the potential for the envisaged hazard to occur. The operational viability of this “out-of-sequence” scenario needs to be clarified. If it proves viable, then a communications-level in-sequence assurance mechanism is insufficient to resolve it. A.3.6 Independent Clearances, Delivered Out-Of-Sequence In the following sequence, two clearances are sent in rapid sequence via data link, with no dependency between the clearances. This is a common operational scenario for the initial data link sector under the operational context stated. The clearances arrive out-of-sequence, and are operationally acknowledged in their order of arrival (i.e. out of sequence). A.3.6.1 Sequence No. 16/02/2016 ATSU Aircraft 1 Uplink “cleared [point]” 2 Uplink “climb to and maintain direct Comment to TRS164/TW/3/3/43 - Draft 0.4 17 TRS164/TW/3/3/43 No. ATSU Aircraft Comment [level]” A.3.6.2 3 Aircrew receives and Wilco’s climb to [level]. 4 Aircrew receives and Unable’s cleared direct to [point]. Conclusion No hazard arises from the above scenario. The clearances arrive out-of-sequence, and are operationally acknowledged in their order of arrival (i.e. out of sequence). 1) Because the controller did not issue the second clearance under the assumption that the first would be acknowledged, there is no hazard associated with this scenario. 2) This scenario is operationally viable, is not a hazard, but if the underlying in-sequence assurance mechanism forced an abort of the connection because of the out-ofsequence delivery, it would result in total loss of the comm. channel with neither clearance dialogue completed. 3) Loss of the comm. channel may, in itself, result in a hazard that would otherwise not have occurred (ref 0 and 0). The overall conclusion is that operationally benign “out-of-sequence” exchanges can occur with no hazard arising. However, a communications-level in-sequence assurance mechanism that mandated removal of the CPDLC connection if out-of-sequence is detected may in fact cause new problems for the controller and aircrew involved. A.3.7 Independent Clearances, Delivered Out-Of-Sequence, with Latent Message Check In the following sequence, two clearances are sent in rapid sequence via data link, with no dependency between the clearances. The clearances arrive out-of-sequence, due to network congestion causing a media switch or e.g. FANS-1/A internetworking. Both arrive after a two-minute latency check in the aircraft (latency value chosen to allow multi-airspace use, and to match Europe’s operational dialogue warning timer). A.3.7.1 Sequence No. 16/02/2016 ATSU Aircraft 1 Uplink “cleared [point]” direct 2 Uplink “climb to and maintain [level]” Comment to Draft 0.4 18 Conclusion and Recommendations No. A.3.7.2 ATSU Aircraft 3 Aircrew receives climb to [level], with a warning that it is old and he should verify it. 4 Aircrew receives “direct to [point]”, with a warning that it is old and should be verified. Comment Conclusion Any hazard arising from the above scenario is at least partially mitigated by the latent message check. Even if no check for out-of-sequence messages is done, crews will receive an advisory that something needs to be verified if the out-of-sequence message arrives after a given parameter. As a probable cause of out-of-sequence messages is either network congestion or internetworking, delays will be inherent to many cases of out-of-sequence messages, and this will result in a latent message error. This means that if a latent message check is in place, out-of-sequence messages are only a problem if: A. The out-of-sequence messages are dependent upon each other. B. The second message is issued without waiting for an operational acknowledgement to the first. C. The messages are issued, reversed, and delivered within two minutes (or whatever latency check parameter is used). The overall conclusion is that this “out-of-sequence” scenario is not a hazard, because no dependency between unacknowledged uplinks occurs. The scenario also shows that a latent message alerting mechanism can partially mitigate hazards arising from out-ofsequence delivery, if such hazards are established. A.3.8 Independent Out-Of-Sequence Downlinks Cause Abort In the following sequence, two downlinks are sent in rapid sequence via data link, with no dependency between them. The downlinks are detected as arriving out-of-sequence by the communications, are discarded without notification to the user, and the CPDLC channel is removed as a result. Another clearance dialogue is in progress at the time. 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 19 TRS164/TW/3/3/43 A.3.8.1 Sequence No. 1 ATSU Aircraft Comment Uplink “climb to and maintain [level]” 2 Wilco to [level] 3 Downlink “request direct to [posn]” Climb begins Comm system unable to deliver messages in sequence and aborts CPDLC connection. 4 Ground never sees Wilco. Unless LACK is used, Controller does not know if the clearance uplink was received. A.3.8.2 Conclusion Three conclusions can be drawn from this scenario: 1. No hazard would have arisen from the out-of-sequence delivery itself. 2. Choosing a communications-level in-sequence assurance mechanism that results in removal of the communications channel, the above non-hazardous out-of-sequence delivery resulted in loss of CPDLC. If data link is being relied upon for capacity increases, or to allow other safety critical changes, then loss of CPDLC may be a hazard in itself. Likelihood of occurrence is an important parameter to establish as input to any future analysis on this scenario. 3. If LACK is not used on the ground, then a hazard may occur from a communicationslevel in-sequence assurance mechanism that results in removal of the CPDLC channel when out-of-sequence delivery is detected. In this case, the controller would not know whether the clearance had been delivered or not unless a LACK or MAS mechanism was used. A communications-level in-sequence assurance mechanism, a secondary requirement may be LACK / MAS. The overall conclusion of this scenario is that one or more hazard(s) may result a communications-level in-sequence assurance mechanism that results in removal of the CPDLC channel when out-of-sequence delivery is detected, though no hazard arises from the out-of-sequence delivery itself. This should be looked at during the safety assessment. 16/02/2016 Draft 0.4 20 Conclusion and Recommendations A.3.9 Independent Out-Of-Sequence Downlinks Cause Abort (2) In the following sequence, two downlinks are sent in rapid sequence via data link, with no dependency between them. The downlinks arrive out-of-sequence, and result in removal of the CPDLC channel when out-of-sequence delivery is detected. No clearance dialogue is in progress at the time. A.3.9.1 Sequence No. ATSU Aircraft 1 Downlink “Request [level]” 3 Downlink “MAYDAY” Comment Comm system unable to deliver messages in sequence and aborts CPDLC connection. 4 A.3.9.2 Ground never sees MAYDAY Conclusion It is understood that crews would normally not rely on data link alone to notify an emergency. However, in some cases it may be the only appropriate means available. The overall conclusion of this scenario is that one or more hazard(s) may result from a communications-level in-sequence assurance mechanism that results in removal of the CPDLC channel when out-of-sequence delivery is detected, though no hazard would have arisen from the out-of-sequence delivery itself (quite the reverse – this out-of-sequence was beneficial). This should be looked at during the safety assessment. Likelihood of occurrence is an important parameter to establish as input to any future analysis on this scenario. A.3.10 Out-Of-Sequence ADS Downlinks Cause Abort In the following sequence, two ADS downlinks are sent in sequence via data link, with no dependency between them. The downlinks arrive out-of-sequence, and result in removal of the ADS channel when outof-sequence delivery is detected. A.3.10.1 Sequence No. 1 16/02/2016 ATSU Aircraft Comment Downlink ADS Report 1 TRS164/TW/3/3/43 - Draft 0.4 21 TRS164/TW/3/3/43 No. ATSU Aircraft 3 Comment Downlink ADS Report 2 Comm system unable to deliver ADS Reports in sequence and aborts ADS connection. 4 Ground does not see any ADS reports, and must reset the surveillance connection if surveillance data desired. A.3.10.2 Conclusion If the ADS link is used as a primary function, required for e.g. reduced separation or increased capacity, the above scenario needs to be analyzed to see if it is a hazard. Hazards could arise from either the reversed reports, or removal of the ADS comm. channel. The overall conclusion of this scenario is that one or more hazard(s) may result from a communications-level in-sequence assurance mechanism that results in removal of the ADS channel when out-of-sequence delivery is detected, though no hazard would have arisen from the out-of-sequence delivery itself. This should be looked at during the safety assessment. Likelihood of occurrence is an important parameter to establish as input to any future analysis of this scenario. A.3.11 Multi-Aircraft, Dependant Clearance, Tactical CPDLC Scenario In the following sequence, three aircraft are involved in the sequences and CPDLC is being used for immediate separation purposes, with dependencies between the clearances. This scenario is one of those used by OPLINK in definition processes for the ATN requirement and solution. It therefore shows the hazard envisaged by that ICAO Panel. This scenario is not in keeping with the currently stated operational concept for air/ground datalink in congested airspace (i.e. CPDLC for strategic use only, with voice used for tactical communications. A.3.11.1 Sequence No. 1 2 16/02/2016 ATSU Aircraft Uplink “climb to and maintain [FL390]” to aircraft 1 Comment Aircraft 1 now at FL370. From aircraft 2 “request to climb to FL370” Draft 0.4 Aircraft 2 now at FL350 22 Conclusion and Recommendations No. 3 ATSU Aircraft Uplink “maintain [FL350] and await further instructions” to aircraft 2 4 5 7 An alternative response would simply be “Unable”. Aircraft 1: Receives and ‘Wilco’s’ “climb to [FL390]” Uplink “climb to and maintain [FL370]” to aircraft 2. 6 Comment Climb begins. Controller observes start of climb on SSR though no ‘Wilco’ received yet. Does so on the basis of observing SSR climb on aircraft 1, above. Aircraft 2: Receives and ‘Wilco’s’ “climb to [FL370]” Uplink “climb to and maintain [FL350]” to aircraft 3. Climb begins. Controller observes start of climb on SSR, but has not yet received ‘Wilco’. Aircraft 3 now at FL330. Controller does this on the basis of observing climb on aircraft 2, above. 8 Aircraft 2: Receives “maintain [FL350] and await further instructions”” Reverts to voice, but may take action right away on this clearance and go back to 350 prior to contacting the controller. A.3.11.2 Conclusion The potential hazard is found at event 8. This case does not apply to the currently defined operational concept for congested en-route airspace, and assumes a specific response to the downlink event rather than simply “Unable” which would not result in this confusion. For other operational concepts, the following observations (not conclusions) are good starting points for subsequent analysis: This scenario is a hazard if: A. 16/02/2016 The controller does not use voice at any stage, despite the complexity of the situation. TRS164/TW/3/3/43 - Draft 0.4 23 TRS164/TW/3/3/43 B. The aircrew of aircraft 2 does not use voice prior to executing the second clearance. C. The controller uses datalink tactically (i.e. for immediate separation purposes). The potential operational environment envisaged was a holding stack in lower airspace, with CPDLC used tactically. D. The controller issues contradictory messages without waiting for a response to the initial message being contradicted. E. The controller does not clarify the above contradiction via voice. F. The controller issues multiple clearances dependent upon each other, without waiting for confirmation of the clearances. G. No latency check is in place on the aircraft to identify and warn crews of old messages. The overall conclusion is that the operational viability of this “out-of-sequence” scenario needs to be clarified. If it proves viable, then a communications-level in-sequence assurance mechanism appears insufficient to resolve the scenario. A.3.12 Confusing CPDLC Downlink Scenario In the following sequence, two downlinks are received out-of-sequence. A.3.12.1 Sequence No. ATSU Comment 1 Downlink [FL350]” “request 2 Downlink [FL360]” “request 3 Receives “request [FL360] and uplinks “climb to [FL360]” 4 5 16/02/2016 Aircraft Aircraft receives and ‘Wilco’s’ “climb to [FL360] Receives “request [FL350]” Message arrives out of sequence, and controller wonders if aircraft wants to descend now. Draft 0.4 24 Conclusion and Recommendations A.3.12.2 Conclusion The following conclusions can be drawn from the above scenario, as a basis for further analysis. 1. The sequence creates confusion, but is not a direct hazard. 2. In a VHF environment, as in the congested en-route airspace being considered, reversion to R/T is the simple solution (assuming the controller doesn’t ignore the request and leave it to the crew to raise it if it’s an issue). The overall conclusion is that this scenario should be looked at in more detail for other environments, in particular to discover if a communications-level in-sequence assurance mechanism that results in removal of the CPDLC channel when out-of-sequence delivery is detected is appropriate in this case. A.4 Conclusion A.4.1 Missing Scenarios? It’s understood that key, operationally viable, scenarios may be missing from this analysis. If such scenarios are known, they should be included. Inputs are solicited. A.4.2 Out-of-Sequence Delivery a Hazard In This Environment? A.4.2.1 Is it a Hazard? Out-of-sequence messages are not a hazard in the operational context outlined at A.2 above. As long as CPDLC is being used without dependency between an unacknowledged clearance and a subsequent clearance, no hazard is apparent. Just as for voice, the hazard only arises if controllers issue clearances dependent upon each other, without getting an operational acknowledgement first. Just as with voice, use of data link to issue multiple clearances, with an unstated dependency between these clearances, and issuance prior to acknowledgement between them, is not in keeping with the currently understood operational context for data link in congested en-route airspace. Irrespective of the underlying technology and message order assurance mechanisms, such usage is not recommended without additional operational analysis. As for voice, using data link in this way requires different procedures and a new end-end safety analysis to examine the full ramifications of such use. This analysis should include things such as aircrew ability to access and acknowledge messages outside their order of delivery and, if the hazards appear to be limited to specific operational environments or ground systems, restrictions on ground system message issuance prior to previous message acknowledgement. A.4.2.2 Is Automatic Comm. Channel Removal (abort) a Good Solution? Automatic removal of the CPDLC or ADS channel on the basis of out-of-sequence delivery detection does not appear appropriate to this airspace. This solution reacts to a nuisance occurrence (out-of-sequence delivery), by removing the communications channel entirely and forcing aircrew and controller to recover all open dialogues through R/T. This includes dialogues unaffected by the out-of-sequence delivery. 16/02/2016 TRS164/TW/3/3/43 - Draft 0.4 25 TRS164/TW/3/3/43 The result is therefore potentially more of a nuisance, and possibly even a hazard, than the occurrence that triggered it. Moreover, unless manual or automatic reconnection is put in place on the ground, the comm. channel is removed for this flight for the duration of its transit through this and any other sectors of the ATC Center. Manual reconnection will not always be utilized, even if available, and automatic reconnection is an expensive undertaking, which brings new safety hazards to be resolved. In either case, the cost/benefit equation is negatively impacted. The next ATC Center may also be affected, as the aircrew must now logon to this center and may have been expecting automatic hand-off. For this airspace, delivery of the message with clear notification of its out-of-sequence state, to higher-level automation and/or the controller or pilot involved, may be a better solution. This would allow the delivery sequence to be dealt with according to its operational ramifications. A.4.3 Out-of-Sequence Delivery a Hazard in Other Environments? If out-of-sequence message delivery is identified as a hazard in other operational environments: 1. Controlling message sequences within the communications layers appears insufficient. Additional requirements may be necessary to ensure that in-sequence clearances cannot be acknowledged out-of-sequence. 2. Controlling message sequences within the communications layers may be overridden at the higher automation layers through alerting or any number of other requirements either in the same datalink specifications or in other specifications imposed on the receiving systems). This should be checked to ensure it would not result in the same hazard. 3. Controlling message sequences within the communications layers, and removing the comm. channel on any occurrence of out-of-sequence messages appears to be inappropriate. It prevents operationally acceptable sequences from occurring without loss of all communications associated with the link. Local requirements, preventing issuance of messages prior to confirmation that the previous message has been received, may be more suitable if the operational situations driving this requirement are unique to an area or unit. 4. Controlling message sequences within the communications layers, and removing the comm. channel on any occurrence of out-of-sequence messages, may create new hazards if the data link channel in question is lost when being relied upon for e.g. capacity increases or separation reductions. This needs to be analyzed. 5. An alternative, and perhaps more suitable requirement for all environments, may be along the lines of: As a general principle messages should be delivered in sequence. If they cannot be, out-of sequence messages shall be delivered and notified to the recipient as out-ofsequence. Such a construct would appear to address all the scenarios outlined in this paper. 6. 16/02/2016 A key consideration in defining the solution is to establish the expected rate of occurrence. This should be done for FANS-1/A as a first step. Draft 0.4 26