4 Air-Ground Data Link Communications in Air Transport – Networking Simulation for Intelligent Transportation Systems

4
Air-Ground Data Link Communications in Air Transport

The current evolution of the civil aviation industry shows a drastic increase in data exchanges between on-board and ground systems. These data are related to safety, eco-friendliness and economic purposes. The overall set of solutions, including the communication system and the applications, is known as the aeronautical data link. Regarding the considered airspace, different communication systems can be used. Some of these recent systems, such as VDL (VHF Data Link), are based on the line-of-sight links between aircraft and ground stations, thus limiting their deployment to the continental domain. In oceanic areas, satellite-based systems are proposed as the main solution for future aeronautical data link communications known as AMSS (aeronautical mobile satellite service). Both systems are intended to support very different types of services with mobile nodes. In this context, traffic characterization, communication architecture and protocols have to be explored and validated. These are the fields in which a simulator becomes handy, allowing the validation of techniques and algorithms.

4.1. Introduction

4.1.1. Context

Aeronautical communication embraces a wide spectrum of usage, from the passengers’ desire for on-board Wi-Fi with Internet connectivity and airline data collection for the cost efficiency of aircraft operation, to safety-related communication between the pilot and the controller for managing the air traffic. All these applications require wireless communication means between the aircraft and the ground, with different Quality-of-Service requirements. Moreover, the environment and conditions in which these communications take place vary widely during the flight: from an on-ground low-speed and dense area on the airport runway to a high-altitude environment at high speed, known as the en-route airspace. The simulations described in the following sections consider those technologies that are dedicated to en-route airspace for safety of life communications, mainly supporting the communications taking place between a controller and a pilot [BEN 13].

Currently, these communications are mainly based on voice analog radios, providing a quite intuitive yet error-prone communication path between controllers and pilots: misunderstandings, inefficiencies and errors in executing clearances, with no automation possible. The allocated spectrum for these communications includes both VHF band from 118 to 137 MHz for continental areas and HF band in remote and oceanic airspace.

The steady increase in aircraft traffic demand all over the world pushed the air traffic industry to look for safety and efficiency improvements. They will be achieved, inter alia, through the use of digital communication technologies, which enable the increased quality and efficiency of the communication path and a higher level of automation.

For several years now, the International Civil Aviation Organization has been working on the development and deployment of the Aeronautical Telecommunication Network (ATN). This worldwide internetwork will gather all the air traffic management stakeholders together, facilitating the sharing of operational information and supporting of near real-time applications for the control of air traffic. This is to say that aircraft are part of this internetwork and should be provided with digital air-ground telecommunication means in the different airspace.

For this purpose, VHF band is a good candidate for supporting these communications with achievable high availability, low delays and high throughput in continental areas. After a technology selection process, it is the VHF Data link mode 2 technology that has been elected as the air-ground sub-network for continental areas and is currently deployed in several regions of the world. VDL mode 2 is the subject of the first simulation model presented in the following sections.

For oceanic and remote airspace, satellite-based communication seems the only technology to provide the Quality-of-Service suitable for the above-described applications and is the subject of the second simulation model to be described hereafter.

4.1.2. OMNeT++

OMNeT++ [OMN 16] is a discrete event simulation system based on C++, which mainly focuses on communication networks and distributed systems. This is an open-source and research-oriented framework. It enables large-scale simulation with hierarchical models. A discrete-event simulation is a chronological sequence of the occurrences of events. This approach requires an event list to be maintained, insertion into and deletion from it to be enabled, the simulation clock to be handled and utilities to generate random numbers from common probability distributions to be provided. Varga [VAR 08] gives detailed information on OMNeT++ and compares it with other frameworks dedicated to network simulations.

4.2. Continental air-ground data link communications and VDL mode 2

4.2.1. Communication system

VDL mode 2 uses the same protocols as those of X.25 at the interface between the aircraft and the ground station, although no requirements apply to the supporting ground wide-area network. VDL mode 2 should provide a reliable connection-oriented network service between an on-board ATN router and a ground ATN router. An additional connectionless service is also provided at the link layer level but has currently no standardized use. Air-ground connectivity is provided through several ground stations, building a mobile network that manages mobility through handing the aircraft over the different stations across its radio coverage. The same radio channel is operated by different ground stations, easing the transition from one station to the other. Ground stations may operate several channels, in which case tuning parameters are provided by the ground stations. The mobile airborne stations will hand over from one station to another by applying the so-called “make before break” paradigm (soft handoff). Managing these handoffs in an efficient way is of prime importance to maintain the connectivity between air and ground routers. An aircraft is required to be able to manage handoffs on its own; however, options are provided for the ground to perform handoffs or to require aircraft to do so, for example, to manage channel load. The latter is also essential to maintain an acceptable Quality-of-Service, especially for maintaining low transit delays.

From a physical layer viewpoint, VDL mode 2 is operated in the aeronautical VHF band and more precisely in the upper channels of this band, between 136.900 and 136.975 MHz (see ICAO Annex 10 vol. 4 [ICA 07]). VDL mode 2 uses the same channel spacing as that of the voice channel, namely 25 kHz channels, and a differential eight-phase shift keying modulation that provides 31,500 bps per channel. Channel coding consists of a Reed–Solomon block coding, with each block being interleaved to spread error burst. Transmission thus consists of a frame with a start of transmission identifier and the length of the transmission to allow the de-interleaving stage to perform its job. This frame acts as a container for the link layer protocol frames that may be grouped into a single access to the channel. The maximum length of a transmission is limited by the transmission length encoded on 17 bits (131,071 bits). A bit scrambling stage with a fixed initialization value is performed before the data are provided to the modulation stage. Additional modulation/demodulation optimization techniques are used to enhance bit error rate. The requirement here is a BER of 10−4 at the output of the physical layer inside the intended coverage.

Channel access follows the CSMA p-persistent rule: for a transmission to take place, the channel is first sensed for idle/busy state. If busy, the station will persist in listening to the channel to wait for it to become idle. Each time the channel is tested idle or has just become idle, the station transmits with probability p and waits for a slot time with probability 1p. After each time slot, the process is started again, until the transmission is performed, or the maximum number of transmission attempts is reached, or the maximum channel access timer expires.

The link layer protocol is a derived version of ISO protocol HDLC, renamed Aviation VHF Link Control. Key characteristics include the use of a modified selective reject frame for selectively rejecting and acknowledging frames intended to minimize the number of unnecessary retransmissions. Acknowledgments are delayed to allow grouped acknowledgments and piggybacking of the acknowledgment into an information frame. Information frames are sent with a transmission window of 4 to accelerate the transmission in case of bursts of data (e.g. several segments of data). No priority, neither within transmission nor between transmitting stations, is defined. On top of the AVLC protocol, an 8208 network protocol allows large data that would not fit into the maximum frame length to be segmented and several virtual circuits inside a single link layer connection to be multiplexed. Flow control may also be achieved here by delaying acknowledgments.

From the viewpoint of link layer connection management, ground stations announce themselves through the sending of an identification frame containing a protocol parameter to be used as well as the DTE address of the reachable routers. Mobile stations are expected to listen for these identification frames to discover the available ground stations and establish the initial link. Handoff and channel load management requires both air and ground to gather information on the peer station. Signal quality measurements on each received transmission and a few timers allow the stations to acquire a reasonably good knowledge about the surrounding other stations to try to manage handoffs in an efficient way. On the ground side, stations sharing the same knowledge are said to belong to the same VDL mode 2 ground system. When the conditions require the aircraft to perform a handoff or if the current ground station requires so, the aircraft will establish a link layer connection with a new station before disconnecting the old link and reestablish all the necessary virtual circuits. Handoffs will happen in regions where radio coverage of at least two ground stations from the same ground system overlaps. There is no requirement for these ground stations to be synchronized for channel access. Handoffs between two separate ground systems are treated as the first link and require an explicit disconnection of the old link. Optionally, the handoff may require the mobile station to retune its radio on another channel. Different deployment scenarios exist in a multi-frequency operation.

4.2.2. Dimensioning parameters and bottlenecks

As explained in the previous part, the VDL mode 2 architecture covers the functionalities provided from the physical layer to the first subpart of the network layer. Of course, the number of dimensioning protocols parameters and bottlenecks is potentially high, considering the relevant layers, particularly in a wireless communication environment. Considering outgoing packets from an end system, a first bottleneck is met in the DLS (Data Link Service) sublayer with the AVLC protocol. In order to ensure point-to-point reliability with flow control, AVLC uses a sliding window with a default size of four frames. Hence, as shown in Figure 4.1, the packets have to be potentially enqueued until previously sent frames are acknowledged. This point is particularly relevant in the ground station, where several DLE (Data Link Entity) may be present, that is, one for each connected aircraft. In the VME (VDL Management Entity) sublayer, each LME manages the AVLC connection between an aircraft and the ground station. Hence, for a given traffic load generated by the upper layers, AVLC parameters have to be tuned in order to ensure an efficient flow control while avoiding congestion in queues. The important parameters are:

  • – the window size k;
  • – the delay before ACK T2;
  • – the maximum number of bits in any frames N1 (default: 8,312 bits);
  • – the maximum number of transmissions N2 (default: 6).

The delay before retransmission T1 is computed and updated during the different connections as a function of several parameters. Notably, the TD99, that is, the observed transaction delay (from application layer to application layer) for 99% of packets is one of these parameters.

The MAC sublayer is also driven by a set of parameters that have to be efficiently tuned. As explained in the previous part, this sublayer is based on the CSMA p-persistent protocol in order to prevent collisions between frames sent by the different nodes.

The main relevant parameters are:

  • – the probability p to transmit if the channel is idle (default: 13/256);
  • – the interaccess delay timer TM1 between two attempts (default: 4.5 ms);
  • – the maximum number of access attempts M1 (default: 135).

Figure 4.1. VDL mode 2 layers and entities

Considering the asymmetric topology given by a group of aircraft covered by a single ground station, it has to be verified that the protocols and different mechanisms operate in a fair way. For instance, the mean waiting time in the queue that feed the MAC sublayer has to be approximately identical in the aircraft and the ground station.

Finally, the physical layer also includes important parameters. In our context, mainly dedicated to CSMA and AVLC protocols, we consider the maximum length of the physical frame (131,071 bits) that allows several MAC frames to be aggregated. Furthermore, it has to be underlined that the channel capacity (31.5 kbit/s) at the physical layer also represents a potential bottleneck.

4.2.3. Simulation model

The main goals of the simulation model are to assess the performances of VDL mode 2, considering the CSMA p-persistent and AVLC parameters, and to provide pedagogic tools to students. As shown in Figure 4.2, the studied model represents a geographical zone covered by a single VGS (VHF Ground Station). The number of visible aircraft is one of the model parameters.

Figure 4.2. VDL mode 2 topology and node models

The model includes two types of node: one VGS (VDL Ground Station) and several aircraft. Another module named medium through which all the messages are sent is also included. Both the ground station (Ground Station module) and the aircraft (Aircraft module) extend the Vdl2node module presented in Figure 4.2. As the behavior of a node is complex, the functionalities of the Vdl2node are split into several submodules.

These submodules are:

  • – Application (app): this module allows the generation of messages from aircraft or from GS application layers. The generated messages have a random length using a uniform distribution between a minimum length of 32 bytes and a maximum length of 265 bytes. And for a single DLE, the time between generated messages has a random value, using an exponential distribution, with a mean value of 40 s (by default);
  • – VDL card: the VDL card module implements all the VDL mechanisms, from the physical layer to the data link layer. It is a complex component that contains several submodules;
  • – Mobility module: this represents the position of a node and provides several useful functions to calculate distances between nodes. Considering the current objectives of the simulation model, node positions are static. They are precomputed and read from a file during the initialization of simulations.

As the VDL card is the core of the model, the following paragraphs offer an insight into its submodules shown in Figure 4.3.

The physical layer modules VDL_rx and VDL_tx, respectively, represent a radio receiver and a radio transmitter. The aims of these two modules are to send (to receive) frames to (from) other nodes through the medium module and to handle collisions of signals.

A MAC module implements the CSMA-p persistent protocol. The value of p can be initialized independently for each node.

Figure 4.3. VDL card model and AVLC module

The AVLC module, presented in Figure 4.3, implements the AVLC communication protocol. It is a compound module. The nDLE submodule represents the AVLC DLE. The Queue module models the queue that stands between the DLEs and the MAC sublayer. This queue plays a very important role in the model. As the CSMA-p protocol senses the channel before sending a packet from the queue, the time spent in the queue may not be negligible. Therefore, a correct management of the queue by DLEs is crucial in order to avoid sending outdated frames and congesting the channel. And, as the radio frames are broadcast to all the nodes, the Switch submodule filters the received frames using the destination address indicated in the frames header. It has to be noted that, as the connection phase is not modeled, there is no module representing the LMEs (Link Management Entities).

The Medium module is designed to broadcast messages sent by nodes and to eventually apply bit corruption and packet losses on the VHF channel. Furthermore, this module sets a propagation delay for each node according to the distance between senders and receivers.

4.2.4. Analysis of simulation results

We illustrate here the different types of potential results with some examples.

Figure 4.4. Sequence charts of MAC and AVLC layers. For a color version of this figure, see www.iste.co.uk/hilt/transportation.zip

The simulation results help us to assess the performances of the VDL mode 2 communication system, considering the number of aircraft covered. How wever, the initial goal of these results is to check if the modeled system and the relevant protocols behave as expected. A method is based on the observation of sequence charts of the simulations generated when the event logging is turned on. This type of result is also useful in cases of pedagogical objectives. Hence, Figure 4.4 shows the sequence charts of MAC and AVLC layers. As expected, when the channel is sensed free, the competing nodes send frames with probability p or wait for a TM1 time slot with probability 1p. As soon as one node obtains the right to transmit, the frame is broadcasted and the other competing nodes will wait until the end of the transmission to continue the process. The AVLC protocol provides flow control and packet loss detection. The chart shows the use of the timers T1 and T2. The first one helps to detect packet loss as when it expires, the sender retransmits the previously sent frames that are still unacknowledged. The timer T2 is used on the receiver side to slightly delay the acknowledgments in order to maximize the probability of sending it with eventual outgoing frames (piggybacking).

Figure 4.5. Number of frames in MAC queue. For a color version of this figure, see www.iste.co.uk/hilt/transportation.zip

Figure 4.5 helps to analyze the fill rate of the MAC queues in both aircraft and ground station. In this simulation, 100 aircraft are present in the VDL cell. In accordance with the results of existing studies and as previously explained, the generated messages in each application have a random length using a uniform distribution between 32 and 265 bytes. And the time between generated messages has a random value using an exponential distribution with a mean value of 40 s. The mean number of frames in the MAC queue is about seven times greater than that in the ground station. This is explained by the fact that each aircraft is connected to one ground station and the ground station is connected to several aircraft, 100 in the considered case. Hence, 100 DLEs feeding one MAC queue are present in the ground station.

Nevertheless, the results presented in Figure 4.6 show that the mean waiting time in the MAC queue is approximately similar for the ground station and the aircraft. This is explained by the fact that as the physical frames are quite long, relative to the MAC frames, their aggregation particularly benefits the ground station in the considered conditions.

Figure 4.6. Mean time in MAC queue. For a color version of this figure, see www.iste.co.uk/hilt/transportation.zip

4.3. Oceanic air-ground data link communications and AMS(R)S

4.3.1. The aeronautical mobile satellite (route) service and Classic Aero

Telecommunication satellites play an essential role in the aviation context because of their ability to provide a worldwide service. Two types of system are currently certified by ICAO: the first, called “Classic Aero”, is based on geostationary satellites (mainly the Inmarsat fleet) and the second is based on the IRIDIUM constellation of low Earth orbit (LEO) satellites. Both systems provide a data link supporting ACARS, FANS and ADS-C services and voice communications. They both provide low data rates when compared to those usually encountered in satellite communications (a few hundred of bits/s to a few kbit/s).

The architecture of the “Classic Aero” system is quite representative of that used for most communication systems employing geostationary transparent payload satellites (also called bent-pipe satellites). The salient features are:

  • – the satellite coverage is very extensive; splitting the service area into zones is necessary from both a performance standpoint (link budget) and frequency spatial reuse (total system capacity). In the case of “Classic Aero”, coverage of one geostationary satellite corresponds to all of the visible area from a geometric viewpoint, about one-third of the Earth’s surface excluding polar regions. This large service zone is then subdivided into regional beams (19 for one INMARSAT 4 satellite);
  • – topologies on forward and reverse links are different. The forward link is the connection between Earth stations (GES, Ground Earth Station) and terminals (AES, Aircraft Earth Station); the return link is the connection between terminals and Earth stations;
  • – for the forward link, the system takes advantage of the broadcast signals from a station to a geographical area (regional or global coverage beam). The physical layer of the system being one broadcast to all, the access method is quite naturally a time multiplex (TDM, Time Division Multiplex). An Earth station transmits on several TDM carriers continuously. Data broadcasted toward a group of aircraft may concern one or more of them; actual reception is based on filtering on the layer 2 address;
  • – for the return link, the available bandwidth is divided into carriers, which must then be used by several aircraft. The radio resource management is MF-TDMA (Multi-Frequency Time Division Multiple Access). Before sending data, each aircraft must identify both the appropriate carrier and time slot. One time slot accommodates a burst.

It is notable that access techniques in the Internet and multimedia geostationary satellites systems are designed on the same principles, even when the data rates are not comparable (several hundreds of Mbit/s). As an example, DVB-S2 implements TDM for the forward link and DVB-RCS2 implements MF-TDMA for the return link.

4.3.2. Dimensioning parameters and bottlenecks

Considering the architecture of “Classic Aero”, the design and dimensioning of the forward link is rather straightforward. The capacity of a given carrier is set by the link budget; the needed number of carriers is determined by the total number of active aircraft within a beam. A simple queuing model allows for delay and congestion analysis. Conversely, the reverse link uses an access method, whose performance analysis can be tricky. MF-TDMA supposes that an aircraft identifies a time slot on a radio frequency carrier before sending one burst. Two access methods are implemented:

  • – a random access similar to S-ALOHA. This random access is of course used for network entry and the corresponding signaling but also for data transmission. The corresponding physical channel is called R for Random;
  • – a deterministic access. A signalization loop allows an aircraft to apply for a transmission capacity to the Earth station and obtain the allocation of a time interval on a carrier for data transmission. The corresponding physical channel is called T for TDMA.

The coexistence of these two access techniques for data transmission is justified by the delay induced by the geostationary satellite hop (about 250 ms). In the case of small data volumes, random access reduces the latency despite the lower efficiency. The downside is that as soon as the data volume to be transmitted becomes more consistent, the probability of data loss by collision and consequently the probability of retransmission may lead to degraded performances.

Figure 4.7. R and T channels access procedure for Classic Aero return link

The return link for “Classic Aero” relies on 11-byte data blocks (SU, Signaling Units) as a format for burst construction by the MAC sublayer. One block can be accommodated within one R channel burst. The MAC sublayer must determine whether one data block from the LLC sublayer should be transmitted over the random access physical channel R or using the deterministic access physical channel T. The decision is based on a simple threshold: the MAC sublayer switches to deterministic access as soon as the volume of data to be transmitted exceeds 33 bytes or three blocks. Figure 4.7 illustrates the signaling process for capacity requests and T slot allocation (note the retransmission timer tA8).

The main issues when designing the system are to ensure the random access technique runs in stable mode and to verify the T channel capacity is suited to that allocated to the R channel (distribution of carriers between the two physical channels). The main metrics to be investigated are:

  • – the random access channel R total load G;
  • – the random access channel R utilization S;
  • – the transmission delay for SU blocks over the R channel;
  • – the utilization of the T channel;
  • – the transmission delay as measured in the LLC sublayer.

The maximum delay observed in 95% of the cases is a system characterization driver.

4.3.3. Simulation model

The simulation model focuses on the analysis of the return link within one beam. Access to R and T channels is simulated with packets sent in radio bursts accommodated in each time slot. The information carried by the forward link is not broadcast but sent from point to point with a delay simulating the one induced by the satellite hop. The number of active aircraft is a simulation parameter; the ability of OMNeT++ to dynamically instantiate objects is thus exploited to change the network load. Figure 4.8 shows the appearance of the interface after loading the model.

The communications are point to point (aircraft to Earth station or opposite direction). The traffic model is similar to the one developed in the VDL Mode 2 model. And with a similar approach, modules of traffic generation and logical link management are instantiated in the Earth station at each entrance of an aircraft in the network. The model is a specific development and does not make use of model libraries like INET; for example, the addressing process relies on 3-byte aircraft identifiers as defined by ICAO (International Civil Aviation Organization).

Figure 4.8. AMSS simulation model. For a color version of this figure, see www.iste.co.uk/hilt/transportation.zip

4.3.4. Analysis of simulation results

The primary objective of the simulation is to enable the analysis of the operation of the access layer and to establish a balance between the R and T channel capacities. The approach is shown in Figure 4.9.

Figure 4.9. R and T channel dimensioning

Analysis of the S(G) trace (channel utilization vs. total channel load) sets the limit operating point of the random access. The results are very close to those of the theory of S-ALOHA access, with no control loop (e.g. by changing the back-off parameters). The limit operating point is then used to determine the corresponding traffic intensity by the curve G(iat) (total channel load in function of the mean interarrival time of messages generated by the application layer). A parametric study is then conducted to determine the number of carriers in T format (TDMA) necessary in order to get T channel utilization close to 1 at the limit operating point. Performance in terms of delay may then be deduced.

4.4. Summary and further work

Random access techniques are currently experiencing a resurgence of major interest thanks to the introduction of signal processing techniques like SIC (Successive Interference Cancellation). The major contribution of these techniques is to enable the retrieval of collided packets and therefore greatly improve performance. In the context of aeronautical satellite communications, the proposed standard enacted as part of the IRIS project [IRI 13] is based on the use of the E-SSA access method (Enhanced Spread Spectrum-Aloha). OMNeT++ is a very suitable tool for studying the performance of such systems, in which the characteristics of the mobile radio channel have a significant impact (distribution of signal powers at receiver input in particular).

Furthermore, random access techniques may also be driven by several parameters similarly to CSMA p-persistent in VDL mode 2. And here again, OMNeT++ is very useful to study the performance of the system under different conditions, considering how the parameters are tuned.

The presented models can of course be improved, particularly by including simulated aircraft trajectories. However, the main driver for further developments will be to build a unified framework for the considered systems (VHF and Satcom data links) and new architectures in order to be able to address the present and future research challenges. We can mention as an example the vertical handover in the presence of heterogeneous communications systems, where on-ground network interconnection and aircraft on-board router designs interact in order to fulfill the reliability, availability and delay performance objectives of ICAO.

4.5. Bibliography

[BEN 13] BEN MAHMOUD M.S., GUERBER C., LARRIEU N. et al., Aeronautical Air-Ground Data Link Communications, ISTE and John Wiley & Sons, 2014.

[ICA 07] ICAO, “Annex 10 to the Convention on International Civil Aviation, Volume III Communication Systems (Part I Digital Data Communication Systems, Part II Voice Communication Systems)”, 2007.

[IRI 13] IRIS, “ANTARES Communication Standard Technical Specifications”, IRIS-AN-CP-TNO-612-ESA-C1, Issue 1.0, September 2013.

[OMN 16] OMNeT++ Discrete Event Simulator, available at: https://omnetpp.org/, 2016.

[VAR 08] VARGA A., “An overview of the OMNeT++ simulation environment”, Simutools ‘08 Proceedings of the 1st International Conference on Simulation Tools and Techniques for Communications, Networks and Systems & Workshops, p. 60, 2008.

Chapter written by Christophe GUERBER, Alain PIROVANO and José RADZIK.