Scalable Mobile Ad-Hoc Networks
Nowadays, the uninterrupted availability of communication networks is self-evident for the vast majority of people. However, natural disasters like the flood in Germany in July 2021 have shown that the existence of a reliable communication infrastructure can not always be presumed. Mobile Ad-Hoc Networks, in the remaining referred to as MANETs, do not require any fixed infrastructure, wherefore they are considered to be highly flexible and disaster-resistant. They can particularly be employed to establish mission-critical communication systems in case the conventional infrastructure-based networks collapse, which makes MANETs highly attractive for public authorities’ security and safety institutions.
In ad-hoc networks often messages originated from one node have to be distributed in the entire network for network control and maintenance. Moreover, safety critical messages have usually to be spread among all nodes. These kind of broadcasts build up a non-negligible amount of the overall network traffic and can cause performance bottlenecks. Node movement and fast fading channels that particularly emerge for vehicular and flying ad-hoc networks (VANETs and FANETs) cause frequent disconnections and rapid topology changes, which exacerbate the necessity for often broadcasts that common ad-hoc network routing protocols rely on. Overall, it has to be stated, that MANETs currently suffer from a poor scalability.
Related information theoretic research has shown that distributed cooperative communication (DisCoCom) on the physical layer, where several standalone nodes send simultaneously, can enable an approximately linear scaling behaviour, but did not consider impairments and constraints of practical systems. In fact, ideal cooperation is presumed, where each sending node’s contribution can be perfectly considered. However, real-world systems suffer from imperfections due to hardware limitations. Besides, specific impairments arise due to aggregating several standalone nodes to a virtual transmit cluster, such as multiple carrier frequency and timing offsets.Transmit diversity schemes, e. g. Space-Time Block-Codes (STBCs), allow for a comparably low-complexity implementation of cooperative communication. Collisions, provoked by the simultaneously sending active nodes are not harmful, but purposely exploited. Nonetheless, adaptations are necessary to utilize them for distributed setups. We utilize this approach to enhance the performance of broadcasting in practical MANETs. In a multistage (or hierarchical) distributed cooperative broadcast scenario one node starts to send. Surrounding nodes, that were able to successfully decode, join transmission and support in distributing the message. Decoding success can be assumed e. g. if the cyclic redundancy check (CRC) of a forward error correcting (FEC) scheme is fulfilled or if the signal-to-noise ratio (SNR) is above a certain
threshold. Nodes, that were able to decode retain active, i. e. send repeatedly, until the complete network is covered, wherefore the number of transmitting nodes is gradually increased for each broadcast stage. The growing number of TX typically enables to cover larger distances, which can be valuable in a distributed long haul multiple-input single-output (MISO) scenario, where several nearby nodes, i. e. a virtual TX cluster tries to reach a distant receiver.
Distributed Cooperative Communication (DisCoCom) Systems
Distributed cooperative communication (DisCoCom) and distributed spatial multiplexing (DisSpaMux) have been theoretical constructs whose benefits for the scalability limitation of MANETs could be shown by information theoretic work. However, they were practically limited, i. e. the maximum number of supported nodes was limited due to the lack of a suitable physical layer implementation. E. g. the famous Alamouti code only allows for two active nodes while other comparable encoding schemes are either limited to a fixed number (which is only slightly higher) of senders as well or come at a cost of significant rate loss. With our research we addressed major issues with respect to the physical layer and were able to overcome these limitations. Accordingly, we can present a physical layer implementation that basically
- supports an arbitrary number of transmit nodes that are simultaneously active,
- allows for a full rate (rate one) transmission,
- does neither require any channel state information at the transmit side (CSIT) nor information about the remaining nodes’ position or a feedback link,
- achieves full diversity,
- is robust against transmitters in deep fade,
- is insensitive to multiple carrier frequency offsets (CFO),
- is robust against multiple timing offsets (TO),
- profits from multipath propagation,
- is computationally efficient (no maximum-likelihood (ML) decoder required, a simple iterative decision-feedback-equalizer (DFE) is sufficient) and that
- enables the use of an arbitrary number of transmit and receive antennas per node that in turn can be leveraged to achieve a spatial multiplexing gain.
Basically, we were able to answer following research questions:
- What is a suitable space-time block-code (STBC) for DisCoCom and DisSpaMux? Which adaptations to this STBC are necessary?
- What impairments arise due to accumulating several nodes to a virtual TX cluster? Can these impairments be mitigated?
- Can the proposed system be demonstrated? (See section about demonstration system)
- What is the impact of employing multi-antenna nodes?
- How can the performance be further enhanced only relaying on decentralized local node-specific knowledge?
- How can the efficiency be increased?
To conclude, fundamentally our proposed communication systems facilitate significantly more reliable, accelerated and cost-efficient broadcasts that in turn can become a key enabler towards better scaling MANETs.
We built a MANET demonstration system (demo system) that allows to study the behaviour of our proposed cooperative communication systems for practical networks. Our demo system is based on software-defined radios (SDRs) and characterized by a highly flexible and scalable architecture. To the best of our knowledge, it is the first of its kind.
Subsequently, we briefly introduce the most relevant aspects together with an exemplarily measurement scenario and result. Furthermore, we provide a video, where we present our demo system in detail.
Our main objective has been to design a scalable and modular architecture that on the one hand allows to arbitrarily add nodes and that on the other hand retrains the computational effort at the controlling PCs manageable. To this end, we decide for a network based solution that leverages edge computing. Fundamentally, we employ Ettus Research B210 SDRs that possess two decoupled RF chains, wherefore each SDR can basically map two network nodes. We make use of this property to ease the system setup and and to minimize the necessary amount of devices in order to keep the complexity low. Since we however do not want to investigate the behaviour for co-located, but distributed nodes, we connect each SDR TX or, respectively, RX output to appropriate antennas with stands. These are mounted on magnetic platelets, which allows us to flexibly modify the exact position and by that the topology of the emulated MANET. All SDRs are connected via USB to either a Raspberry Pi computer (RasPi) or a PC. Lastly, we utilize a reference clock to mimic a medium access control (MAC) protocol. By relying on a pulse-per-second (PPS) and 10 MHz reference signal we can ensure that all TX nodes send approximately at the same time and the approximately same carrier frequency. It is crucial to highlight, that a strict synchronization is not mandatory because our proposed communication systems enable a sophisticated mitigation of multiple imperfections. The system model is visualized in the next Figure.
The next figures depict a global view of the demo system and a close view of the SDRs together with the controlling Raspberry Pis (RasPis), the reference clock as well as network
switch. Thereby, each colour (below the antenna stands) represents a
broadcast stage. More details about the specific measurement scenario are explained
in the subsequent section.
Exemplarily Measurement Scenario and Result
We study the propagation of a message during a typical multi-stage broadcast scenario. For that, we rely on the topology that is depicted above. First, we consider the green node to initiate a broadcast and observe the decoding behaviour for a remote RX node (white node). Next, we suppose that the orange nodes could be reached, wherefore they become potential transmitters as well. Thereafter, we assume that the blue nodes join and finally the red ones. Accordingly, we consider four broadcast stages in total. It is worth stating, that we do not check the decoding success of each node, but just consider the coloured progress for the message propagation. For each broadcast stage, we calculate the outage probability pout at the remote RX which we derive based on the bit error rate (BER) of each arriving packet, whereas we arbitrarily select 20% as failure threshold.
To point out the benefits of distributed cooperative communication (DisCoCom), we compare the attainable pout with those obtainable with concurrent communication (each node basically sends the exact same signal without any diversity encoding) and a typical TDMA communication (only the best positioned TX of a stage is active).
The next figure summarizes our measurement results for aforementioned scenario, from which it immediately becomes apparent, that DisCoCom empowers the most reliable transmission among the considered communication types. Concurrent transmission might be beneficial and in fact is more advantageous than a purely TDMA-based transmissions. Nonetheless, it does not achieve the same low outage probability of DisCoCom, which is a perfect demonstration for the benefits of latter.