Phase 1 Part 1

In this study a MEO system was evaluated considering a constellation of 8 satellites and 7 possible ground station locations at different climatic areas. The outage capacity offered at 5 different scenarios was evaluated keeping the latency less than 150 ms due to round trip and trying to have an availability of 99.9%. The ground terminals were considered with two kinds of antenna classes at 1.8 m and 3.5 m. Four RF scenarios were simulated one at Ka-band (reference) and three at Q band (advanced system) without considering any diversity, one using orbital diversity and one for site diversity. The fifth scenario considered wavelengths at optical range using the site diversity technique.

For the system design at Q-band, firstly an identification of available bandwidth at Q-band from ITU regulations was performed. It was found that 5 GHz of available bandwidth exist for satellite downlink. Since there was an available bandwidth of 5 GHz for the downlink keeping the number and size of onboard antennas the same with the Ka-band case, an 833 MHz of bandwidth at user beam was identified. Therefore, for fair comparison, the same mass on satellite and power consumption was considered. Moreover, for the system evaluation at RF, first an optimization on the PHY layer was performed to derive the ModCods at Ka and a second optimization to derive the ModCod table at Q band. More particularly, at Q band for the PHY layer two carriers were considered at each beam to reduce the noise power. For the calculation of capacity, realistic time series generator of attenuation due to atmospheric phenomena was used taking into account the spatial and temporal correlation of the attenuation factors.

From the capacity results it was found that on average there is a 100% increase in system throughput at Q/V band in comparison to reference scenario in clear sky for the same power consumption and mass on board. Moreover, in Q/V band the mean system throughput at clear sky is close to 15 Gbps for the ten user beams. However, in the single link case the 99.9% availability is not always reassured.

Using orbital diversity again the 99.9% availability is not reassured for all the stations but there is an increase in both the availability and capacity. Finally, with site diversity the availability is always on target. Here it must be noted that orbital diversity can be already implemented with the existing O3b system since every ground station has already two antennas.

Finally, for the optical system in most places more than 6 ground stations are required for achieving the target availability of 99.9% and close to 80Gbps can be delivered on ground.

The goal of WI is the design of new advanced signal processing techniques to enhance the throughput of fixed and mobile satellite communications. In more detail, the objective will be driven by two approaches: 
- The use of polarization as an additional degree of freedom has been popularized and has become a realistic new way of MIMO communications. In this WI, more than two polarization planes will be used to increase the overall performance. On the one hand, in fixed links, the long coherence time and the preexisting feedback links can be exploited to provide the Gateway with Channel State Information to adapt the transmission according to the channel magnitudes. On the other hand, the high uncorrelated polarization channels can be used to provide multimedia communications in mobile devices improving the reliability and the latency. 
- Full Duplex has been presented as a new reliable schema to increase the spectral efficiency by transmitting and receiving at the same time at the same frequency. Knowing the transmitted patterns may be used to suppress the interference added to the receiver path. In this WI, new techniques will be presented to outperform the performance of the system by using this approach. 

This work item deals with a literature survey on the adoption of Network Coding (NC), differentiated on the basis of the OSI layer where network coding is adopted. Therefore, this document is divided into three parts, as follows:  Network coding adopted at the application layer (or just below); Network coding applied at the transport layer (or just below);  Network coding applied at the network layer (or just below). The purpose of this document is to survey the different applications of network coding explaining why it is used and how it is used with a particular attention to multipath scenarios and satellite case studies.

The second part of this work contains a detailed description of the selected simulation scenarios by the team and a description of related issues in terms of the layer where NC should be used and NC compatibility issues (ROHC, IPsec). Finally, benchmark techniques are presented and a description of the implementation phase of simulators (3 scenarios) is provided together with some results and comparisons with benchmark schemes.

The third part of this activity contains a survey of currently available simulation tools for network coding and discuss pros and cons.


Simulations results and final considerations are also provided that have permitted to highlight the advantages in using network coding.

The objectives of this WI is the design of an end-to-end functional system architecture for an integrated satellite-terrestrial network based on ICN/PSI (Information-Centric Networking/Publish-Subscribe Internet) to support the delivery of M2M/IoT services, investigation of the impact of M2M/IoT traffic on multiple access schemes, test-bed validation of ICN/PSI architectures over enhanced satellite multiple access schemes, and optimization of critical network functions of ICN/PSI architectures over enhanced satellite multiple access schemes.

The investigations will consider various M2M/IoT scenarios where integrated terrestrial-satellite networks offer a significant advantage. The investigations will assess the performance in terms of the reliability, scalability, and signaling overhead, identifying their corresponding tradeoffs and the impact of M2M/IoT traffic characteristics.

Phase 1 Part 2

The performed work and the obtained results can be separated in two parts:


Nanosatellite part: a simulation platform has been developed to simulate a Nanosatellite constellation network mimicking a Web Browsing Data Exchange service between hosts located in rural areas and Internet Servers linked to the Internet. The simulator is based on the software Network Simulator 3 (NS3) and includes: a scenario module, to allow arbitrarily setting different parameters related to the network design in order to be able to simulate different scenarios; a nanosatellite movement module, to realistically simulate nanosatellite movement following a defined orbit; a traffic module, to implement a sort of Delay Tolerant Networking (DTN)-HTTP Web Browsing communication; a DTN module, to implement the characteristics of the DTN paradigm needed to realize a communication in this kind of network, such as the store and forward mechanism and the relay action to transmit among heterogeneous portions, and a personalized and light version of the Bundle Protocol. A source routing algorithm based on the Contact Graph Routing (CGR) algorithm has been proposed to guarantee a reliable communication considering the limited available resources. A channel model has been developed to simulate both Nanosatellite/Ground Station and Nanosatellite/Nanosatellite communications, taking into account the local environment effects such as attenuation, fading, Doppler effect. A performance evaluation campaign has been performed in order to test the proposed routing algorithm and to quantify the obtained performance for the considered service in this network. Different scenarios have been simulated changing the number of Nanosatellites and of Ground Stations, defining a proper set of orbital parameters and of satellite link budget parameters. The obtained performance has been collected in terms of Average Delivery Time and Computational Burden on-board nanosatellite. The results show that the proposed network offers a possible solution to give a basic Internet access to rural areas with low performance but an affordable cost especially for remote areas or underdeveloped countries.



UAV part: a simulation platform based on the software S-NS3, a satellite network extension to NS3 platform, has been developed to simulate a scenario composed of different UAV swarms which deliver data to a Ground Control Station (GCS) through a Geostationary (GEO) satellite. In particular, this scenario involves different swarms delivering M2M traffic to a remote sink application connected to a satellite gateway. Each swarm is composed by a single cluster-head (CH) and multiple UAVs (E) that may possibly also act as routers (R). The CH encapsulates a CoAP proxy module: UAVs gather data through on-board sensors and send it to the “client side” of the proxy. Multiple CHs will compete to contemporarily access the Random Access (RA) satellite channel. A first set of simulations has been performed in order to numerically evaluate and then compare the considered scenario to a MQTT-based one. In particular, the completion time of CoAP and MQTT data exchange has been analyzed for different values of transmission window size (NSTART), obtaining that CoAP provides a lower completion time than MQTT and lower values for larger NSTART values. A second set of simulations has been performed in order to evaluate the average delivery delay. The results show that it is not affected by the NSTART because of the low considered load traffic conditions and the chosen limited number of transmitting nodes. Moreover, mobility slightly affects the delivery delay whereas it tends to increase the packet drops. An alternative ARQ protocol based on a variable-length sliding window mechanism has been proposed. This modified CoAP version outperforms the standard one. Also a congestion control algorithm inspired to the TCP Friendly Rate Control (TFRC) algorithm has been designed. The performance of the proposed solution (CoAP/sb-TFRC) has been compared to the MQTT/TCP solution based on TCP Hybla congestion algorithm. The benefits are evident in terms of average throughput, application-layer goodput, Round Trip Time (RTT), and MAC queue size. Finally, a satellite channel affected by rain fading has been modelled and the obtained results show that the impact of the rain attenuation is non-negligible. The stability of both Hybla and sb-TFRC is preserved, although in the former case the penalty appears to be slightly more evident.

Cryptographic direct-sequence spread spectrum (DSSS) modulation is currently being considered for updating the telecommand standard for space applications. To achieve a better immunity against pulsed jamming scenarios, DSSS can be combined with advanced channel coding schemes. However, an essential prerequisite for reliable channel decoding is accurate synchronization. The first goal of this WI was therefore to develop, analyze and evaluate of symbol and frame synchronization techniques for advanced channel-coded DSSS telecommand links under jamming conditions. The performed work comprises the selection of potential techniques and the identification of the trade-offs in terms of computational complexity, implementation cost and error performance. Computer simulations have been performed, to validate and compare the proposed solutions.


As a first candidate frame synchronization scheme we considered the ALG-0 algorithm. It involves a minor modification of the detection procedure currently described in the existing standard. The algorithm is based on hard symbol decisions and has a very low complexity. Compared to the frame synchronization algorithm specified in the current Consultative Committee for Space Data Systems (CCSDS) recommendation, we considered a longer start sequence and relaxed the condition for declaring synchronization. The performance of this algorithm was investigated in the presence of jamming, and it was shown that the frame synchronizer can be designed such that the overall system's robustness against pulsed jamming is limited by the robustness of the code rather than the synchronizer.


While the ALG-0 frame synchronizer was chosen with the objective to minimize the impact of the introduction of the new synchronization scheme to the existing standard, a possible way to further improve the frame synchronizer performance is to adopt more involved algorithms. Unfortunately, optimum frame synchronization requires jammer state information (JSI) which is not readily available at the receiver. Two practical frame synchronizers that circumvent this problem without decreasing the performance too much have been proposed. Both these synchronizers have been shown to significantly outperform ALG-0.


Finally, we have also studied the performance of symbol timing acquisition procedures under jamming conditions, assuming DSSS with asynchronous BPSK modulation of the PN sequence. Attention has been focused on a first order phase-locked loop (PLL). We have proposed different loop gain selection methods. It was demonstrated that, if JSI is available at the receiver, PLLs with variable loop gains perform better in terms of acquisition speed and tracking accuracy than a conventional PLL with a constant loop gain.


The above mentioned DSSS techniques are physical layer security mechanisms of space links that mostly rely on low probability of intercept (LPI) and low probability of detection (LPD) waveforms. In effect, LPI and LPD waveforms along with signal processing and channel coding techniques can be highly effective against jamming attacks and interception. Their anti-jamming performance can be quantized in terms of the power level

and signal at which the protection is broken. As for interception, security performance can be quantized in terms of breaking the pseudo-random sequence spreading the signal. However, novel attacks and computing power evolution could end up breaking their computationally strong security. In this document, we have introduced the information-theoretic approach to the design of keyless physical layer security for space links.  This method exploits physical characteristics of the communication channel to provide unconditionally secure communications. Moreover, the secrecy level of these information theoretical techniques can numerically quantify secrecy according to information-theoretic principles. This allows the design of unconditionally secure cryptographic coding schemes that can lead to practical schemes adapted to channel, system and mission specific constraints and design goals. 


Specifically, we have studied the applicability of the concept of wiretap coding to space links. To this aim, we reviewed different secrecy criteria that have been introduced in the literature and can be numerically quantized.

We then focused on cryptographic semantic secrecy for which we introduced wiretap code constructions that are available in the literature. In order to analyze its applicability to space links, we developed a novel signal model that includes the relative geometry of the eavesdropper with respect to the legitimate receiver as well as the respective system-level parameters. In our analysis of the infinite-length regime we obtained the secrecy capacity of typical space Gaussian links for different relative geometries and system parameters. On the other hand, in our analysis of the finite-length regime we presented upper bounds of the semantic secrecy showing how communication with secrecy guarantees can be designed trading o_ reliability and secrecy according to mission-specific needs. Our numerical results confirm the potential of information theoretic physical layer security for space channels. Finally, we identified system layered design options for enabling simultaneous operation with additional information theoretic security protocols. Our final remark is to stress once more that the relevance of this direction of research lies on the fact that off-the-shelf transmission codes can be chosen to ensure reliable transmission over the legitimate channel while a new preprocessing layer is added to ensure secrecy.


In a nutshell, the aim of this WI has been to evaluate how the use of multiple simultaneous Spreading Factors can benefit a satellite-based access network system based on E-SSA, and also to design a transmission strategy which can attain maximum performance in the presence of M2M devices powered-up with energy-harvesting devices, thus not having continuous availability of energy.


A MATLAB simulator has been implemented to evaluate the performance of the proposed techniques. This document describes the simulator and also shows and discusses the most relevant results that have been obtained. The work conducted in this WI has established a solid base towards an interesting research area.


There are two key conclusions which have been obtained throughout the work conducted in this WI:


1.     When no Energy Harvesting (EH) is considered at the M2M terminals on the ground, the use of more than one Spreading Factor (SF) is beneficial for the SIC process and improves the overall performance of the E-SSA algorithm. Adding more than one SF in an M2M satellite system based on the E-SSA protocol adds a new source of diversity and improves the overall system throughput.


2.     When considering an EH-aware transmission policy, results show that a technique which dynamically adjusts the spreading factor used in each terminal according to its available level of energy achieves a higher throughput and guarantees a reduced outage probability compared to the case of using a single and fixed SF.


In addition, the following research lines have been identified for future work:


·         When no EH is considered, it would be interesting to find the optimal distribution of SF amongst the population of E-SSA terminals which can provide the highest possible throughput. An analytical study may be carried out in order to determine the optimal distribution of the SFs.


·         For the energy harvesting scenario, further studies on this topic involve a detailed tuning of the transmission parameters (e.g., setting the thresholds of the transmission manager), the extension of the system model to consider other inefficiencies, as well as the consideration of other sources of energy harvesting at the transmitter side.



·         A major area of research would consist in analyzing the performance of the E-SSA protocol in a LEO scenario (with and without EH). Due to their high velocity (of several kilometers per second), LEO satellites introduce a high Doppler effect that leads to a frequency drift. This drift depends on the relative velocity of each single node with respect to the satellite and this fact may have a significant impact on the overall system performance. This frequency drift provokes the next two opposite effects: 1) due to due to different drifts that the gateway observes from each node, the packets transmitted in the same frequency band are no longer completely overlapped, i.e., for packets using the same band, the collisions are less destructive because two interfering burst are rarely completely overlapped, and 2) if guard bands are not properly designed, this frequency drift can provoke collisions with transmissions carried out in the adjacent frequency bands.

The telecommunication companies need to sustain/update the infrastructure, which brings the content from points of presence to the users. However, most of the Internet revenue goes to content providers such as Google and Facebook. At the same time, since some of the links are wireless, the scarce spectrum is another issue. In this WP, we have proposed using mono/multi-beam hybrid satellite-terrestrial networks along with off-line and ICN caching algorithms to substantially off-load the backhaul links.


The results showed that the proposed hybrid satellite-terrestrial architecture is able to considerably reduce the required time for file placement while keeping the cache hit ratio very close to the upper bound. In addition, the proposed multi-beam architecture improves the cache hit ratio since cache feeding can be done based on the global popularity of each beam and is closer to BS content popularity distribution. It was revealed that for a long range of values of file popularity parameters, α, the multi-beam architecture outperforms the mono-beam architecture in terms of the cache hit ratio.  This study showed that the hybrid network can improve the cache hit ratio for specific range of file popularity parameter, α, and cache memory. It was shows that coded caching with optimal memory allocation can substantially improve the cache hit ratio. As a future work, it will be interesting to perform online popularity estimation using learning algorithms to perform real time caching.


The next part investigated four intelligent ICN caching solutions that jointly exploit the advantages of ICN, which include network layer content-awareness and dynamic multicast group creation, and the wide-area broadcasting capability of satellite networks: 1) Opportunistic multicast for proactively caching future chunks of a requested video, 2) Proactive video caching based on video clustering, 3) Partial video caching, and 4) ICN caching and Network Coding.


The first scheme opportunistically exploits the wide-area broadcasting capabilities of satellite networks to proactively cache future chunks of a video that is requested by a user, which has been previously requested by other users and hence is being broadcasted over the satellite network. Validation results were performed for Zipf and non-Zipf popularity distributions (Zipf with exponential cutoff and stretched exponential), showing how the cache hit ratio depends on the aggregate video request rate, the video duration, and the video pool size.

The second scheme considers that videos have been previously clustered into groups, e.g. based on their type and/or user requests, and utilizes information of previous video requests to proactive cache video that has not been seen before at a particular satellite terminal or is new. Results investigated the influence of the cluster popularity and the difference between global and local popularity on the cache performance.


Partial video caching extends the previous scheme by treating different parts of a video differently, based on their relative popularity. Validation results illustrated the gains that can be achieved by partial caching and showed that these gains depend on how accurately the video view time is known.


The study investigated the gains from combining ICN caching with Network Coding, in terms of the reduced data and signaling traffic over the satellite network and the reduced delay, when Random Linear Network Coding (RLNC) is performed by all caches and when it performed only at the original server that made the video available.

 Finally, within the study we extended our testbed that integrates an ICN Publish Subscribe Internet (PSI) prototype implementation and satellite emulation using the OpenSAND tool, with SDN functionality; the testbed was used to validate a subset of the schemes developed within the study.

Precoding for full frequency reuse fixed satellite systems has reached a level of academic maturity and is being considered for implementational studies by operators given the potential gain it offers. Building on some initial works, this study investigated the use of precoding in full frequency reuse mobile interactive multibeam satellite systems. It involved consolidating the channel models for various mobile scenarios - Slow Nomadic, Maritime and Aeronautical - spanning the spectrum of terminal speeds.


Based on the channel models, the outdated channel state information at the transmitter which creates impediments to the use of available precoding techniques was investigated.  Due to the fading nature of the satellite communication which equally impacts the desired and interfering signals, it was shown that the precoder matrix is sensitive to the outdated amplitude and not to the phase of the channels. With a focus on unicast scenario, several precoder designs were considered including the traditional Zero-Forcing, MMSE, those based on SLNR reduction as well as optimal precoders derived from various SINR formulations. To cater to different payload architectures, both sum power and per-antenna power constraints were imposed and practical implementation issue of margins for rate allocation were considered. The performance of the various precoders were evaluated on a INMARSAT type of system and it  was shown that the considered scenario does not suffer from a sum-rate reduction in case a perfect rate allocation is performed.  On the contrary, in case the rate allocation is based on a delayed version of the SINR,  a sum-rate degradation was observed. Despite this performance loss, MMSE precoding can increase the sum-rate a 55 % factor given a certain outage target. In addition, it is observed that for the same margin, precoding has smaller outage compared to the benchmark system. 



Subsequently, a multicast scenario was considered with a geographical scheduling; while MMSE precoding could still achieve gains over the traditional four colour re-use, the gains reduced with the multiplexing order. Finally, the study also discussed different approaches to precoding implementation including on-ground, on-board and hybrid precoding with arguments favouring on-ground precoding if feeder link bandwidth expansion is not an issue.

This project has evaluated the use of Q/V and W bands for future satellite communications. The modelling of the propagation channel, spectrum and regulatory issues are addressed as well as future satellite and earth terminal equipment characteristics. Aeronautical, earth resource down links and point to point professional use applications are evaluated as to their capacity improvements over Ka band. The following are the major conclusions;

·         For operations above Ka band there is no exclusive satellite bands and thus operations will occur in shared bands. At Q/V the portions—(D) 39.5 to 40.5GHz and (U) 48.2 to 50.2GHz and at W band portions-(D) 74 to 76 GHz  and (U) 84 to 86GHz seem most promising for Europe. For Aero systems it is only around airports where co-existence studies would be needed. For larger earth resource gateways these would need coordination in the normal way. For professional terminals if few in number would also need coordination but if uncoordinated operation were required this would need further regulatory decisions. At these Q and W bands it is likely that satellite will receive competition for spectrum from 5G systems in the future.

·         No requirements for Q and W band earth stations on the move are currently in place but these are likely to emerge in the future. It is expected that they will be treated as FSS on the move and need to adhere to FSS regulations.

·         For the Aero application we have shown that the use of Q and W spot beams overlaying Ka can provide increased capacity and are realistic for 2020 and beyond. The use of new flat plate or conformal arrays will enable increases of up to 60% above Ka band and capacities on a 125MHz carrier of 233Mbps per plane which compares with predicted requirements of 175-200Mbps and which could not be attained at Ka band alone.

·         For the earth resource satellite downlink we have evaluated the performance at three locations and shown that Q and W band enable 54% and 132% greater capacity than at Ka band using predicted earth station and satellite equipment. Diversity was also investigated and shown to have the ability to increase availability for smaller antennas or reduced satellite power with all availabilities being in excess of 99.9%. Diversity at W band could especially  combat the effects of clouds but has improvement for rain alone,

·         For the two way Professional mesh links, the use of Q/V band and large antenna size (2-3m) enables an increase in the average throughput with regards to what can be achieved using Ka band. However even using very low SNR modcods the availability is one order of magnitude lower. The use of W band with the assumptions used does not lead to a sufficient availability and does not constitute a viable alternative.


·         A comprehensive evaluation of the receiver ACM loop was performed using Q and W band predicted time series for both one-way and two-way scenarios. It was shown that in all cases, the average spectral efficiency (throughput) could be optimised with ACM margins of 0.8dB or less over the duration of a rain event. These values will reduce further when optimising for transmissions that include a predominant clear sky profile.  The use of variable or adaptive margins could have some advantage at Q band but on the limited data available at W band there was low correlation on the fade slopes which indicate that there would not be much to gain. Further study with measured data is recommended for the future.

The aim of the study was to investigate the use of Ka-/Ku-band  LEO satellites to extend and complement terrestrial networks, so as to provide broadband connectivity to urban or rural areas not served by means of a terrestrial infrastructure. The Team, as jointly agreed with ESA, considered LTE Rel. 13 as a reference for the radio interface and assessed the feasibility of LTE waveforms and PHY/MAC layers procedures, in order to identify which adaptations would be required.

In the first phase of the study, the Team identified the scenario to be analysed and the main channel impairments to be considered, i.e., large Round Trip Time (RTT) delays, large Doppler shifts, and Doppler rate. Based on a thorough review of the allowed architectures in LTE Rel. 13, the Team identified two potential system architectures: i) a satellite Relay Node (Sat-RN) architecture, in which a satellite-enabled RN provides an on-ground LTE cell; and ii) a satellite eNodeB (Sat-eNB) architecture, in which the satellite-enabled entity on-ground is a traditional LTE eNB. It is worthwhile highlighting that in both cases the terrestrial user link (access link) is provided by means of the traditional Uu LTE air interface. Thus, the main difference between the two scenarios resides in the backhaul link, which depends on the type of satellite-enabled entity. In agreement with ESA, the scenario taken into account for the feasibility assessment has been the Sat-RN scenario, in which the on-ground LTE RN is connected to the satellite by means of a Un air interface.

In the second phase, the Team focused on reviewing the Un air interface of LTE Rel. 13. In particular, the main addressed areas have been waveform design and PHY/MAC layer procedures. The Un air interface is almost identical to the traditional Uu interface and, thus, it is based on SC-FDMA in the uplink and OFDM in the downlink. The Team, in agreement with ESA, assumed a FDD framing structure with normal Cyclic Prefix (CP) length as the reference waveform and analysed the impact of the large Doppler shifts (due to the large relative velocity between LEO satellites and fixed RNs) on the subcarrier spacing. By means of geometrical considerations and through mathematical assessment, it has been demonstrated that, in Ku-band, as long as the RNs are capable of estimating their position relative to the satellite with a maximum location error of 4 km, the Doppler shift can be compensated at the gateway and, thus, no modifications would be required.

With respect to the PHY and MAC layers procedures, the Team analysed the impact of the large RTT (approximately 16 ms) on the different timers involved in the procedures. In particular, the focus has been on the backhaul link, since the terrestrial access link is a traditional LTE cell for which no modifications are needed. In particular, the following challenges have been identified:

·         The LTE Random Access procedure is a 4-steps procedure in which two different timers are present: i) the RA response window, which can be as long as 13 ms; and ii) the contention resolution timer, which can be as long as 64 ms. It can be noticed that, while the latter is significantly above the RTT and does not pose any challenge, the former is shorter than the RTT and, thus, proper solutions are needed.

·         The RA procedure with RNs comes into play only during the RN attachment procedure. This is a 2-step procedure in which the RN first connects as a normal user to gather the parameters so as it can operate as a relay and, then, it actually connects as a relay. Although the RA response window is below the RTT, the Team observed that, since the RNs are fixed and the satellites orbits are known, ad-hoc network deployment solutions can be envisaged in order to perform the RN attachment procedure, which only comes into play at network setup.


·         In the HARQ procedure, ACK/NACK feedbacks are based on a 4 ms periodicity both in the uplink and in the downlink, which is significantly below the large RTT with LEO constellations. Thus, the HARQ procedure cannot be implemented as is and proper modifications are required.

Phase 2 Part 1

In the framework of 3GPP standardisation, RAN studies and activities are now providing significant results and the first release of New Radio (NR, i.e., 5G) specifications has almost been finalised. In addition, a new 3GPP study item in the 3GPP framework has been recently started for 5G non-terrestrial networks (NTN). In this Study, in order to understand the feasibility of a 5G-based satellite system, the applicability of the most promising 5G techniques proposed within 3GPP has been assessed in the framework of Satellite Communications, which requires an in-depth analysis of both satellite-terrestrial networks and State-of-the-Art 5G techniques and PHY/MAC procedures. To this aim, based on a survey ton he use cases and scenarios from both the 3GPP and NetWorld2020 fora, two services have been identified to be addressed: i) extreme coverage at lower rates (eMBB scenario); and ii) local area IoT (mMTC scenario). Several architectures and deployment options have been reviewed and analysed, which are different in terms of access link (direct to the user or through on-ground relays) and satellite payload (transparent or regenerative). Focusing on transparent payloads, the direct to the user access link has been selected for the mMTC scenario (in particular, focusing on NB-IoT) and access through relays has been selected for eMBB services. The following aspects have then been addressed during the Study:

·       The main impairment related to NR PHY have been identified and evaluated. Given the standard GEO satellite distance from Earth, even assuming perfect alignment between transmitters and receivers along the path, delays are generally much larger than those related to typical terrestrial systems, and so they seem to be not compatible with existing typical cellular requirements, like very low latency communications. This is the major problem to deal with. Moreover, due to the different channel conditions with respect to terrestrial communications, NR over satellites has to be as flexible as possible in many settings, e.g., assuming OFDM waveforms, different subcarrier spacing, subframe duration, and cyclic prefix length have to be supported. Different coding schemes and multiple access mechanisms, along with low SNR at the receiver, have to be as adaptable as possible. Performance evaluation shows that, for any combination of impairments introduced by the channel, it is possible to achieve NR requirements when different performance metrics are taken into account, by optimal tuning.

·       We have considered the candidate waveforms proposed for NR and, in particular, focused on the filtered OFDM (f-OFDM) as the most promising candidate due to its high level of spectral discrimination and the flexibility it provides. The impact of the nonlinear satellite amplifier on the performance of f-OFDM has been investigated. While f-OFDM achieves an OOBE suppression of about 150dB in a linear channel in comparison with about 30dB achieved by OFDM, its performance degrades to that of OFDM in a conventional TWTA channel without predistortion. This is due to the high level filtering applied to f-OFDM. This degradation can be rectified when predistortion is used with the conventional TWTA wherein an OOBE suppression greater than 100dB is achievable. However, we note that this optimum performance is only possible when the OBO is greater than or equal to the PAPR of the f-OFDM signal, whose value is higher than the original legacy OFDM.

·       Ww assessed the impact of the typical satellite channel impairments, (large path losses, delays, and Doppler shifts), on the NR Air Interface and PHY/MAC procedures for the eMBB scenario with transparent payload. In particular, on the one hand the Random Access and Timing Advance do not pose any peculiar issue thanks to the presence of on-ground Relay Nodes, which allow the UEs to connect to what they see as a traditional terrestrial NR network. On the other hand, the HARQ procedure is deeply impacted by the large delays and these values might require a significantly larger number of parallel HARQ processes, which also affects the soft buffer sizes. Different solutions are proposed to keep the number of processes and buffer size under control. In addition, the RA procedure shall be performed by the Relay Node at its start-up. Although the large delays can pose a significant challenge in this phase, since the RNs and the satellite are in fixed locations/orbit, ad hoc network sart-up procedures are proposed.


·       We investigated the possible implementation and use of the NB-IoT standard for Low Power Wide Area Network (LPWAN), defined by 3GPP for better serving IoT use cases, in NTN. More specifically, the study aimed at analysing the feasibility of integrating NB-IoT with Low Earth Orbit (LEO) satellite communication systems. The main challenges have been investigated and possible solutions have been also proposed, empowering the utilization of the NB-IoT terrestrial standard within satellite channel. In particular, timing constraints, differential doppler shift and Extended idle-mode Discontinuous Reception (eDRX) have been analysed taking into account a direct access Satellite-User Terminal. While eDRX and timers have been found compatible with NB-IoT, the differential doppler shift is a source of potential degradation. As a consequence, some solutions based on the geographical users position or on the estimation form the downlink channel, have been proposed. 

Satellite gateway virtualization opens up new paradigm to push more functionality from dedicated hardware to general purpose commodity platforms and to separate the software instances from the hardware they operate on. The general purpose hardware platform could be either implemented locally (as part of the existing gateways), or some applications scenarios (such as start-up systems with small allocated bandwidth to be processed) the use of remote cloud computing could be envisaged.

In principle, the gateway function virtualization could start from the physical layer at the gateway front-end pushing the modem functionality into the cloud. This allows for duplicating the satellite ground station front-end into multiple transmit/receive nodes.

The performance gain and cost-benefit of this concept is highly influenced by the use case and the maturity of the technology to connect multiple Tx/Rx nodes and perform centralized signal processing.

This technical note investigates the use of multiple nodes with the aim of either reducing the system capital and operational expenditures and or to tentatively to increase the achievable rates. We consider the use of multiple antennas of both at the satellite payload and on-ground in a variety of scenarios and frequencies.

Unlike in the GSO case where additional GWs are used for diversity or for aperture increase, multiple GWs are needed in NGSO to ensure seamless handover between multiple space assets. Such GWs are endowed with the ability to track the satellites making them premium for wider proliferation. The activity focused on eliminating the tracking requirement and consider the use of a larger number of antennas to counter the pointing loss. Based on realistic beam cuts and parameters, the work summarized the gain-complexity trade-off, indicating how large number of small-sized untracked antennas can counter pointing loss.

The aim of this work item is to investigate the theoretical bounds of multibeam satellite communication systems and evaluate practical coding schemes that do not require full CSIT. The theoretical analysis reveals that beam cooperation and rate splitting ideas have to be exploited to outperform conventional orthogonal schemes, such as FDM. To approach the theoretical bounds to some extent with a reduced feedback overhead, we have introduced NCRS and SC-SCD coding schemes. Unfortunately, SC-SCD only gives a satisfactory performance when users can be ordered in a natural way from the strongest to the weakest. Interestingly, in the rest of cases, NCRS outperforms SC-SCD. Accordingly, NCRS stands as the most promising scheme. In order to measure the improvement of NCRS in a realistic multibeam satellite communication system, we have performed system-level simulations in a 2-color frequency reuse scheme. The gain of 2-color NCRS with respect to 4-color FDM can be in the order of 20% at low SNR and gets barely unnoticeable for high SNR regimes. Clearly, the presence of a significant background interference, which cannot be addressed by NCRS, is drastically limiting the gain. Aside from investigating scenarios with uniform traffic demand, the WI 3 overviews resource sharing strategies to address non-uniform traffic demand. It has been shown that the throughput of a hot spot can be significantly increased by pulling resources of 6 adjacent beams and allowing receivers to use SCD strategies. Finally, it has been investigated the impact of practical implementation aspects on the performance. Especial emphasis has been given to the design of receivers that can recover two non-orthogonal signals in presence of symbol misalignment. At the expenses of increasing the complexity, the proposed receiver can cope with symbol asynchronous conditions. 

Recent studies, clearly highlighted that new techniques, e.g., interference management, aiming at improving system performance (throughput, flexibility, and robustness,) have a fundamental impact on the design and selection of resource allocation algorithms and, in particular, on scheduling policies. Moreover, it was shown that physical layer (PHY) performance adopting such new techniques are deeply affected by resource assignment policies. This Study aimed at evaluating the performance of scheduling algorithms on the Forward Link with multiple traffic classes at Medium Access Control (MAC) layer for broadband Satellite Communications (SatCom) systems also taking into account PHY features, constraints, and cross-layer requirements, with both Time Slicing and multicast precoding.

Time Slicing aims at reducing the complexity of FEC decoding at the receiver. Two QoS traffic classes and related requirements have been envisaged, real-time and  non-real-time services, and several scheduler architectures have been reviewed from the literature. A two-layer scheme was proposed that is suitable for time slicing in DVB-S2/S2X which takes several aspects into account, such as IP class of service, PHY-layer conditions, BBframe generation, and Time Slice (TS) structure. Starting from the Magister SNS3 simulator for DVB-S2, several modifications have been implemented to take the TS structure and the QoS differentiation into account allowing the simulator to schedule the User Terminals (UTs) so that a certain UT uses the same TS in the cycles. The scheduler provides results in terms of average delay, PLR, distributions of delay, jitter, and Jain fairness index. A variant of the Proportional Fairness (PF) scheduler (PF with exponential rule, PF-ER) was adopted to achieve a fair resource allocation of the TS resources among the MODCODs, taking also deadlines into account. The PF-ER and Round Robin (RR) solutions were compared, showing that they achieve similar results for the highest QoS1 service class. On the other hand, PF-ER outperforms RR for the QoS2 class for all performance indexes. Finally, it was shown that PF-ER has a superior performance in terms of the fairness index.

Focusing on an aggressive full frequency reuse (FFR) scheme in a GEO HTS multibeam satellite system, several aspects have been assessed with respect to multicast precoding, i.e., a precoding solution in which the same precoding matrix is applied to all of the symbols in the same codeword, meaning that the precoding matrix is constant over a time frame, and multiple users are properly selected and multiplexed in the same codeword. The performance of multicast precoding has been shown to be deeply impacted by the similarity metric that is used to group together (cluster) the users to be multiplexed in the same FEC codeword and it is based on the selection of a random reference user to be grouped with the closest users around it. Two metrics have been evaluated for the clustering procedure: i) Euclidean distance; and ii) distance in the channel coefficients space. The latter has been shown to provide better performance in terms of average spectral efficiency, minimum intra-cluster SINR, and CDF of the intra-cluster SINR standard deviation. In addition, the performance has been also compared to that of a 4-colour frequency reuse showing that multicast precoding provides a valuable gain in the performance. Although providing significant gains in the overall system rate, it has been observed that there are often critical scenarios in which some users experience a lowest SINR with precoding w.r.t. the absence of precoding techniques. To circumvent this issue, and observing that this aspect is mainly related to the location of the clusters to be served across the system beams, i.e., a scheduling issue, a Geographical Scheduling Algorithm has been designed and assessed, showing a significant performance benefit w.r.t. traditional random scheduling.


In addition, the forward packet scheduling has been assessed in the above system. The scheduling and the precoding design are closely coupled with each other, making it very challenging to provide a joint optimal solution that can be implemented in practical systems. On the other hand, future broadband satellite systems have to be capable of accommodating heterogeneous services and guarantee their corresponding uneven Quality of Service (QoS) requirements. As a consequence, a novel cross-layer scheduling algorithm was proposed which takes into account the PHY framing together with the modulator and precoding functionality combined with system constraints imposed by QoS requirements in upper layers. In particular, the proposed scheduler aims at: (i) Grouping the users within a frame according to similar channel conditions; (ii) Minimizing the inter-beam interference by scheduling users within adjacent synchronous frames according to orthogonal channel conditions; and (iii) Giving priority to delay-sensitive real-time traffic, while ensuring an acceptable throughput to non-real-time packets.

Transport Layer Security (TLS) solutions with HTTPS enhance end-user privacy but negate almost all middlebox functions operating above the transport layer, including caching, content/protocol optimization, and security filtering tools. This study investigates two solutions for middlebox support to achieve end-to-end behavior similar to that of HTTPS, that involve TLS splitting and entail different tradeoffs in terms of deployment, performance, and privacy. The first solution is keyless-TLS, with which only handshake messages requiring a server’s digital signature are forwarded from the middlebox to the TLS server. Keyless-TLS is transparent to end-users, but requires modifications at the TLS server, and is used by major CDN providers, including Cloudflare and Akamai. The second solution is based on DANE (DNS-based authentication of Named Entities), and allows binding of domain names to certificates belonging to middleboxes. Unlike keyless-TLS, DANE requires modifications at the TLS clients rather than the TLS servers. With TLS splitting, end-users have no control over the path between the middlebox and the TLS server. Although, in general, this is considered a disadvantage, in integrated satellite-terrestrial networks it can be turned into an advantage, since it allows higher flexibility to select a security solution for the path that includes the satellite link and benefit from the wide-area broadcasting capabilities of satellite networks. The main results of this study include the following: Keyless-TLS requires one message exchange with the TLS server for establishing a new connection between a client and a middlebox, as opposed to three messages when vanilla TLS is used. The gains in reduced control message delay are higher in satellite networks compared to terrestrial networks, due to the former’s higher RTT, and in scenarios involving mobility between different middleboxes. With DANE, if the client is connected to the DNS server through a fully terrestrial path, then no signaling messages traverse the satellite link. On the other hand, a disadvantage is that the TLS server cannot directly measure its content’s statistics. Additionally, certificate revocation with DANE, which is completely controlled by the TLS server, can exploit the satellite’s broadcasting capabilities when the revocation message must be sent to multiple DNS servers. When an application layer security solution is used in the path between a middlebox and the TLS server that includes a satellite link, then the transmitted content can be opportunistically cached by middleboxes located close to the recipient middlebox and content requests can be batched, or even aggregated. However, application layer solutions introduce new privacy threats: a malicious entity can distinguish whether two transmissions concern the same content, as opposed to TLS where a separate key is used for each transmission.  

Phase 2 Part 2

Security protocol part:


In this WI we first of all performed a thorough analysis of the security vulnerabilities of current spread spectrum-based multiple access satellite systems, including physical layer mechanisms such as E-SSA (Enhanced-Spread Spectrum Access).


Then, we identified a number of potential security mechanisms that could endow our multiple access satellite system with the CIA security triad (Confidentiality, Integrity and Availability). Potential schemes were either based on information theoretical security or computational security. We chose the mechanism for the downlink known as Asynchronous Spread Spectrum (USS), which is a keyless secure message transfer protocol that is used in combination with (classical) cryptography to provide CIA security. Hence, it provides computational security based on realistic computational assumptions for both the legitimate system and the attackers.


We pre-designed a physical layer keyless protocol that we termed as dual USS (DUSS) for the uplink, which is to be used with USS in the downlink for keyless secure transfer of required cryptographic keys at MAC layer. Hence, our proposal not only is a secure keyless physical layer multiple access protocol but also solves the key distribution problem for the cryptographic primitives at the MAC layer.


Specifically, we proposed a novel secure satellite system model, which we used to derive the attack models to define DUSS operation and integration with E-SSA. We also identified the relevant security performance metrics. In particular, we showed that a secure multiple access protocol requires that the multiple access (stable) capacity should no longer be only measured in terms of energy or noise as it is the case in current systems. Now it is also required to relate the capacity with the computational and/or hardware properties of the system. Further, we have also proposed novel mechanisms for attack detection so that the system can operate in secure or not secure mode. This may allow for optimal throughput achievability by trading-off throughput and security guarantees.


We have also shown that the final design can be pretty flexible as we identified several system parameters and mechanisms that can be subject of optimization. This also facilitates the standardization of our proposal.


IoT Access Scheme part:


In this part of Work Item 4.1 we considered the impact of multiple jamming strategies on an IoT satellite uplink based on a random access protocol E-SSA by using simulation and analysis. Two key components of the receiver, the correlator and the decoder, have been studied to derive the performance of the Successive Interference Cancellation (SIC) scheme affected by the presence of jamming attacks. Particular emphasis has been given to the correlator, which primarily affects the overall performance of the SIC scheme.

Different types of jamming can be used to attack a communication system and the focus of this study is on smart jamming where the jammer emulates the transmission of a regular user. Two possible implementations of smart jamming have been compared: regular packet jamming (REG) and continuous preamble jamming (CP). It has been shown that REG jamming has very limited impact on the system operation unless a disproportionate use of transmission power is done, which makes the jamming terminal easily spotted out. On the contrary, CP jamming with the same average power as the other system users has been shown to be very effective in disrupting the system operation while maintaining the jamming terminal undercover. The effect has been illustrated by showing the growing number of correlator output peaks depending on the fraction of REG/CP jamming packets in the system as well as by the corresponding Receiver Operating Characteristic. Every correlator peak trigger the receiver decoder with a delay depending on the total length of the packet so that a high number of peaks due to jamming packets wastes an impressive amount of system resources at the receiver.

In another stage, the effect of increasing the chip rate (depending on the TFI mode) has been considered. It has been shown that the number of SIC iterations depends strongly on the TFI mode by a precise analytic relationship and performance analysis. An insufficient number of SIC iterations reduces its effectiveness and the resulting system good throughput. Too many iterations represent just a waste of system resources which must be avoided. The resulting high-rate performance has been reported in terms of false-alarm/missed-detection probability plots versus the detection threshold used by the correlator.

Finally, a jamming detection technique has been proposed, which finds the relationship between the correlator output and the fraction of jamming packets in the system. Even though the local number of correlator output peaks may not be representative, a long-term average provides significant insight in the detection and quantification of an ongoing jamming attack, as illustrated by the simulation results.

This study addressed the application of non-orthogonal multiple-access techniques (NOMA) to those satellite communications for which a significant imbalance in the link quality of user terminals can be expected. The SINR (Signal-to-Interference and Noise Ratio) imbalance can be mainly caused by the coexistence of different types of terminals, possibly with different antenna sizes, without excluding other reasons such as traffic asymmetry and relative location under the satellite beam radiation diagram. This link SINR asymmetry can be exploited to outperform the performance of orthogonal access schemes under different rate metrics, paying special attention to fairness in the service provision.

In the case of the forward link, an initial analysis on the exploitation of PD-NOMA by the current standard DVB-S2X was made. A first assessment of the framing implementation for PD-NOMA was presented with the aim to understand the framing requirement of PD-NOMA and its possible implementation in the current standard. An initial evaluation of the performance of DVB-S2X MODCODs in combination with NOMA was followed by a system level study for a single beam scenario. Gains in the order of 20% can be expected for the overall sum rate and the strong users sum-rate with respect to the orthogonal case, as long as the received SNR gap is above 10 dB and the weak users have a relevant provision of rate. Results were also extended under a beam-free approach, for which a given user is not necessarily served by its dominant beam in a multibeam coverage. Even with identical terminals, PD-NOMA can offer an improvement in the minimum rate of the coverage around 8-15%, more significant for more unbalanced traffic demands across beams.

As to the return link, the goal was also to understand whether the theoretical performance gains of NOMA could lead to any practical performance advantage in the satellite return channel. In order for NOMA techniques to offer a considerable gain, two scenarios were considered: (i) Heterogeneous coexistence of broadband satellite terminals with a large imbalance in transmit EIRP, and (ii) Heterogeneous coexistence of IoT satellite terminals with a large imbalance in transmit EIRP. The first scenario can be due to a mixture of different services such as fixed VSAT and aeronautical services, and conventional demand assignment is applied to the chosen pairs by the scheduler, under the condition of large transmit imbalance. Numerical results report gains of the average spectral efficiency up to 91% for EIRP gaps of 14 dB. In the second scenario, E-SSA is used as random access mechanism, and gains up to 61% in the maximum throughput are achieved, again for a 14 dB EIRP between both classes of user terminals.

When it comes to the synchronization of the two superposed non-orthogonal signals in the return link, the strong user (class-1) parameter estimation is hampered by the presence of the weak terminal (class-2) signal in the observation, whereas the class-2 signal parameter estimation is disturbed by the residual interference due to imperfect cancellation of the other signal. Different parameter (time delay, frequency offset, carrier phase and amplitude) estimators were assessed, concluding that very small residual class-1 parameter estimation errors is a prerequisite. Otherwise, it is not possible to achieve the minimum class-2 parameter estimation MSE performance that is needed for proper class-2 data detection. In consequence, this first synchronization phase needs to be followed by an estimation refinement after data detection, otherwise the required accuracy of all the ancillary tasks when extracting the weak signal may be jeopardized. Finally, it was also seen that class-2 users should preferably use sufficiently large bursts with many pre-amble, post-amble and pilot symbols and a Root-Raised Cosine transmit filter with a sufficiently high roll-off factor.

In the Working Item 4.3, the use of deep learning on the resource allocation problem for flexible satellite communication payloads has been investigated. In the current multi-beam satellite communication systems, the power and bandwidth allocation are uniform over all beams and therefore they offer uniform capacity at all the beams. However, the demand traffic is non-uniform and by adding flexibility in the payload offers at least the capabilities to allocate different power level at different beams and assign different carriers at each beam and finally allocate non-uniform capacity. There are two main flexible payload architectures the non-beam hopping and the beam hopping flexible payloads. In resource allocation problem for flexible payloads the power allocated and the assigned carriers to each beam must be optimized in terms of a selected metric (e.g. differential capacity, useful sum-rate) between the offered capacity and the traffic demand. In this study, first the two flexible architectures have been studied in terms of link budget parameters. Then the deep learning methods have been reviewed in terms of their applicability to the resource allocation problem. In particular, solutions have been identified and sketched using supervised, unsupervised and reinforcement learning techniques. Only auto-encoder like neural networks have been developed and tested and only for the case of non-beam hopping flexible payload. In particular three solutions have been developed with neural networks which try to reproduce the input to the output adding a specific layer developed in Tensorflow which transforms the outputs of the neural networks to offered capacity. The first solution, that has been developed provides the weights on the power and the probabilities to assign each carrier at a beam while the second one calculates also the probabilities of assigning a beam to an amplifier.  Although, at a small total traffic demand the developed neural net solutions after the hyper parameter tuning, has resulted to a better root mean square error than the uniform allocation at high total traffic demand the solutions provided by the developed neural networks were not good. Searching through the neural network it was found that the gradients of the losses with respect to weights for neurons, which were connected to carrier allocation were close to zero. Therefore, a third solution has been developed, in which carriers to different color schemes are assigned. It has been proved that provides better results compared to uniform allocation solutions. Finally, the neural networks were examined in terms of generalization. After generating random traffic patterns using uniform distribution for training and validation sets, it has been found that while the neural networks have been trained using the training set the error to the validation set has been also decreased thus leading to the conclusion that the neural networks  could be generalized.

In this study, we first developed a Massive MIMO (M-MIMO) over satellite simulator and validated it against data provided by ESA. Second, we studied the comparison between a M-MIMO scheme based on a direct radiating array and a reference precoding over fixed beams scheme (PoFB) based on an array fed reflector. Third, we carried out a feasibility study on the M-MIMO scheme, focusing on the CSI acquisition problem and on precoding and scheduling solutions. M-MIMO presented sum rate gains between 12% and 45% with respect to PoFB, which were mainly attributed to its flexible power distribution capability. Focusing on the use of DRAs, the combination of a minimum distance algorithm with a simple MRT provided a good low complexity solution, though large improvements can be expected by the use of MMSE in the case on non-uniform user distributions or large available RF power on the satellite. Finally, we showed that the CSI acquisition problem can be solved either performing an array calibration and a DOA estimation in the uplink, or relying on a beam based M-MIMO scheme. This last is a very promising solution since it exploits the main benefits of both PoFB and MMIMO schemes, with a performance penalty below 5% compared to free M-MIMO precoding.

The aim of this work item was to study and develop efficient beamforming optimization techniques using per-element power constraints suitable for active antennas with a large number of elements. For this purpose, we have used two approaches to solve the optimization problems. These approaches are based on semidefinite relaxation (SDR) based solutions and alternating direction method of multipliers (ADMM) based solutions. These techniques specifically address spot beams, shaped beams, or a combination of these that are used in satellite communications. Moreover, an optimization scenario and its parameters, defined and agreed with ESA at the start of the work form the basis of these investigations. The performances and the computational efficiency of the selected algorithms have been compared with each other using the agreed scenarios. The SDR techniques, although very popular in the literature, are not well suited for satellite communication systems. This is due the fact that the computational complexity of the semidefinite program solvers is very high. Moreover, the SDR based techniques are very sensitive to the tightness of the constraint parameters. However, our investigations show that the ADMM based solution results in an excellent performance, low co-channel interference, and much lower computational complexity for large active antenna arrays. Therefore, such techniques are well suited for satellite communications.

We use cookies to facilitate navigation and improve your experience across our website. By clicking "Accept", you will be storing these cookies.