Elsevier

Computer Networks

Volume 180, 24 October 2020, 107358
Computer Networks

Licensed shared access for 5G: Which auction mechanism to choose?

https://doi.org/10.1016/j.comnet.2020.107358Get rights and content

Abstract

Licensed Shared Access (LSA) is a complementary solution allowing Mobile Network Operators (MNOs) to use another incumbents frequency spectrum after obtaining a proper license from the regulator.

Using auctions to allocate those LSA-type licenses is a natural approach toward an efficient use of spectrum, by controlling the incentives for MNOs to declare their true valuation for the spectrum and allocating it to those who value it the most. A specificity of LSA licenses lies in the interactions among buyers, due to possibly overlapping coverage areas, this allows for allocating the same spectrum to several MNOs.

In this paper, we review the existing mechanisms taking into account such radio interference constraints, propose new ones, and compare their performance. We show how to increase the revenue, while maintaining truthful-telling, of all-or-nothing auction mechanisms by introducing a reserve price per bidder. We also investigate extensions of those mechanisms, namely when the management of interference among base stations is more subtle than partitioning base stations into groups of non-interfering base stations. For each mechanism, we show how to optimize a trade-off between expected fairness, expected revenue and expected efficiency by carefully working with groups and reserve prices. Simulations suggests that the extension of those mechanisms may lead to increase an indicator combining allocation fairness, social welfare and seller’s revenue by more than 20% compared to the ones without the extension.

Introduction

Accommodating exploding mobile data traffic is among the greatest challenges for fifth generation (5G) networks [1]. Dealing with that traffic indeed requires an optimal utilization of spectrum, but currently some holders of a licensed spectrum (e.g., militaries, satellites, some commercial users) do not always use all their frequencies–usage varies with time and geographical location–, hence there is some room for improvement, which has given rise to the proposal of the concept of dynamic spectrum access (DSA) [2].

DSA refers to the situation in where a primary user, who has an exclusive right to use the band, shares his bandwidth with a secondary user. Secondary users must allow the primary user to use his spectrum without disrupting it. For this, these systems typically use cognitive radio [3]: secondary users–Mobile Networks Operators (MNOs) in our context–can intelligently detect those communication channels that are in use and those that are not, and move to unused channels. However, for MNOs this approach is risky because neither the access to spectrum nor the quality of service (protection from interference) are guaranteed.

In November 2011, in order to support the deployments of 5G systems [4], the Radio Spectrum Policy group (RSPG) has proposed a new sharing concept called Licensed Shared Access (LSA) [5]. That concept involves three stakeholders: the incumbent user, the secondary user which is called LSA licensee, and the regulator [2]. Contrary to DSA, under the LSA approach, the secondary user needs to obtain a license from the regulator before accessing the spectrum of the incumbent. The license includes the conditions of sharing, in particular in terms of time, frequency and geographic region. The LSA concept guarantees to the incumbent and to the LSA licensee a certain level of QoS specified in the LSA license. The LSA licensee is typically an MNO, we shall thus use the terms MNO, (network) operator, or LSA licensee interchangeably. Likewise, we shall use the terms regulator, seller and auctioneer interchangeably.

Deploying an LSA system requires the introduction of two new architectural building blocks [6], as shown in Fig. 1: the LSA repository and the LSA controller. The LSA repository is a database which contains information about LSA spectrum bands together with their conditions of sharing. It is controlled by the regulator and the incumbent, and is required to deliver the information on spectrum availability based on the incumbent spectrum use and associated conditions for sharing. The LSA controller resides in the network operator’s domain and controls the access to the incumbent’s spectrum by following the instructions received from the LSA repository. Each MNO has to have his own LSA controller. Several trials of the LSA concept have taken place in Europe1 and have shown its applicability. LSA is now under the final stages of standardization and field validation [7] as regards the technical aspects, but the specifics of how to allocate and price spectrum among several potential secondary users remain open.

The LSA concept involves two major differences with regard to the allocation of 3G or 4G spectrum to operators (which already uses auction schemes). First, the allocation needs to work at a faster time scale, since the availability of LSA spectrum will be changed by the incumbent via the LSA repository, possibly several times per hour, and the regulator has to allocate the LSA spectrum for potential LSA licensees as soon as the incumbent releases his spectrum in order to improve the use of the spectrum. Second, spatial re-usability (MNOs who do not interfere can use the same spectrum bands simultaneously), should be leveraged. We will in particular consider a scenario in which multiple base stations of different operators compete for LSA spectrum at a defined period of time in a particular geographical area; no two interfering base station should be allocated the same spectrum, which is ensured by dividing MNOs into groups as will be detailed in Section 3.

A key objective for LSA is to allocate the spectrum in the most efficient way, so as to maximize the resulting value to the market. Since the LSA ecosystem involves several actors (incumbent and MNOs) with nonaligned objectives, one needs to define allocation and pricing schemes that are robust to manipulation; hence the focus on auctions for that task. To the best of our knowledge, there are only a few research studies on auction mechanism design focused on the LSA context. This paper aims at analyzing and comparing auction schemes introduced in the literature for the specific LSA context under different scenarios, as well as benefiting from more general results on auctions ([8], [9]) to propose alternative mechanisms. To compare mechanisms, we apply the commonly used efficiency and fairness measures ([9], [10]), in addition to the fulfillment of properties such as incentive compatibility (truthfulness) which intuitively means that reporting true valuation as a bid maximizes ones payoff, and individual rationality which ensures participants a non-negative payoff.

The rest of this paper is organized as follows: after summarizing the paper’s contributions, in Section 2 we define what an auction mechanism is and describe some of its desirable properties, while the system model we consider is introduced in Section 3. Section 4 contains the main contributions of this paper: under the assumptions made in the literature, we review some proposed mechanisms, adapting one to ensure truthfulness. We also adapt them to include a per-buyer reserve price set by the auctioneer while maintaining incentive properties, and numerically compare those mechanisms with others from the literature in terms of efficiency, revenue, and fairness. Section 5 investigates the relaxation of a key assumption in the model: while the mechanisms partition the base stations into separate groups and allocate spectrum among groups, we consider allowing overlapping groups (groups still covering all base stations, but not necessarily in a partition). This relaxation may improve efficiency of the allocation but complicates the mechanism analysis (ensuring truthfulness becomes harder). Indeed, the payment of each base station which belongs to the winner group is a function of bids of other losing group(s). When relaxing that assumption, a winner base station could be also in another losing group(s) therefore its bid may impact its final payment. Finally, we provide some concluding remarks and suggest some perspective for future work in Section 6.

The first auction mechanism which was proposed in the LSA context is named LSAA [11]; we have shown that this mechanism is not incentive compatible, and proposed PAM [12], a truthful auction mechanism that outperforms LSAA in terms of revenue and fairness. However, PAM relies on the assumption that players will accept even an infinitesimal portion of the LSA bandwidth. All the mechanisms proposed in the literature specifically for LSA –reviewed in Section 3.5– except PAM rely on the assumption that each base station must belong to one and only one group (for otherwise they are not incentive-compatible). The main contributions of this paper can be summarized as follows:

  • We show how to adapt PAM to a more realistic setting: the regulator sets a minimum fraction α: each player must get at least that amount or he gets nothing.

  • We design two new truthful spectrum auction mechanisms from the non-truthful LSAA, named TLSAA and TLSAA2. We prove that the revenue generated by TLSAA2 equals the second-highest bid and is thus may be an attractive choice from the auctioneer’s viewpoint.

  • We show how to increase the revenue provided by the proposed mechanisms, while maintaining truth-telling, by introducing a reserve price per bidder.

  • We give the regulator more flexibility in the group construction by allowing each base-station to be in multiple groups, and by showing how to adapt the payment rules of the previous mechanisms, when possible, for any group configuration, to maintain truthfulness bidding without modifying the allocation rule.

Section snippets

Auction mechanisms and desirable properties

In this section, we provide the definition of an auction mechanism, and of possible properties (goals) that a regulator may want the mechanism to satisfy. Note that each designer of an auction mechanism may be interested in a particular subset of properties.

System model for LSA auctions

In this section, we instantiate the general auction framework to the specific context of LSA auctions. More specifically, we describe our model for players (here, operators and the regulator) preferences, and explain how the interference among coverage areas is managed, through the definition of groups of base stations. We then describe the general working of an LSA auction scheme and present some auction mechanisms which were proposed in that context.

Proposed mechanisms

This section introduces several alternative mechanisms that we suggest could be applied to auction LSA spectrum. We start by introducing PAMα which is an extension of PAM after introducing a minimum amount α i.e., if a player i gets a fraction αi ≠ 0 then αi must be higher than α. Then we present two new auction mechanism named TLSAA and TLSAA2 which are two extensions of LSAA [11]. Finally we show how to increase the revenue for any monotone and all-or-nothing allocation rule. Note that in

Extensions of the previous mechanisms: A base station can belong to more than one group

In this section, we consider relaxing the assumption made previously by treating the case where the base station (BS) grouping allows a base station to belong to several groups, which should improve the efficiency of the allocation but complicates the mechanism analysis (ensuring truthfulness becomes harder).

Indeed, the assumption of allowing each BS to be a member of only one group may appear to be restrictive because by removing this assumption, i.e. allowing a base station to belong to more

Conclusion

In this paper, we have designed new truthful auction mechanisms aimed at allocating spectrum in the context of LSA. We have also studied the impact of an hypothesis found in all the literature, i.e. ”each base station must belong to one and only one group” on truthfulness and we have extended previous studied mechanisms to the scenario in which this hypothesis is relaxed by finding the corresponding payment rule (if it exists) eliciting truthful bidding. The studied mechanisms have different

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships thatcould have appeared to influence the work reported in this paper.

Ayman Chouayakh: was born in Tunisia, in 1992. He received the double degree from SUPCOM Tunisia and SUPELEC, Gif-sur-Yvette, France, in 2015 (the engineering degree in telecommunications and the M.Sc degree in wireless communications). He is currently pursuing the Ph.D. degree in telecommunications at Orange Labs jointly with IMT-Atlantique, France. His research interests include auction mechanism, Game theory and wireless communication.

References (30)

  • X. Ge et al.

    5G ultra-dense cellular networks

    IEEE Wireless Commun.

    (2016)
  • M. Matinmikko et al.

    Spectrum sharing using licensed shared access: the concept and its workflow for LTE-advanced networks

    IEEE Wireless Commun.

    (2014)
  • A. Amraoui et al.

    Accs Dynamique au Spectre dans le Contexte de la Radio Cognitive

    2 edition de la confrence nationale de linformatique destine aux udiants de graduation et de post-graduation

    (2012)
  • Z. Sadreddini et al.

    Dynamic resource sharing in 5g with lsa: criteria-based management framework

    Wireless Commun. Mob. Comput.

    (2018)
  • E.C. Committee, ECC Report 205, 2014, (Available at...
  • V. Frascolla et al.

    Dynamic licensed shared access-a new architecture and spectrum allocation techniques

    Proc. of IEEE VTC Fall

    (2016)
  • S. Yrjølå et al.

    Licensed shared access evolution enables early access to 5g spectrum and novel use cases

    EAI Endor. Trans. Wireless Spect.

    (2017)
  • R.B. Myerson

    Optimal auction design

    Math. Oper. Res.

    (1981)
  • V. Krishna

    Auction Theory

    (2009)
  • R. Jain et al.

    A quantitative measure of fairness and discrimination for resource allocation in shared computer systems

    CoRR

    (1998)
  • H. Wang et al.

    Spectrum Sharing Based on Truthful Auction in Licensed Shared Access Systems

    Vehicular Technology Conference

    (2015)
  • A. Chouayakh et al.

    PAM: A Fair and Truthful Mechanism for 5G Dynamic Spectrum Allocation

    Proc. of IEEE PIMRC

    (2018)
  • M.J. Osborne et al.

    A Course in Game Theory

    (1994)
  • Z. Li et al.

    Designing truthful spectrum auctions for multi-hop secondary networks

    IEEE Trans. Mob. Comput.

    (2015)
  • I.A. Kash et al.

    Enabling spectrum sharing in secondary market auctions

    IEEE Trans. Mob. Comput.

    (2014)
  • Cited by (2)

    Ayman Chouayakh: was born in Tunisia, in 1992. He received the double degree from SUPCOM Tunisia and SUPELEC, Gif-sur-Yvette, France, in 2015 (the engineering degree in telecommunications and the M.Sc degree in wireless communications). He is currently pursuing the Ph.D. degree in telecommunications at Orange Labs jointly with IMT-Atlantique, France. His research interests include auction mechanism, Game theory and wireless communication.

    Aurélien Bechler: is a data-scientist in Orange, France. He graduated from ENSAI (Ecole Nationale de la Statistique et de lAnalyse de lInformation) in 2011 and obtained a PhD in statistics from AgroParisTech (2014). His main research interests are mathematical modeling, data analytics, game theory and learning algorithms.

    Isabel Amigo: is an associate professor at IMT Atlantique. She obtained a PhD (2013) in computer science from Telecom Bretagne, France, and Universidad de la Repblica, Uruguay, and an electrical engineering diploma (2007) from Universidad de la Republica. Her main research interests are traffic engineering, interdomain QoS, New architectures and networking paradigms, network economics, game theory, and network performance. She has been a post doc researcher between 2013–2015 at Telecom Paristech, Paris, France.

    Loutfi Nuaymi: is Full Professor at IMT Atlantique since Feb 2001. He teaches different courses related to the wireless access technologies: GSM/GPRS/EDGE/UMTS/HSPA/LTE/4G/5G, WLAN and Bluetooth. His research activity is in the domain of radio resource allocation and green wireless networks (pls check research page). He is Senior IEEE Member. He is responsible of the ”Network and Mobile Services” (Réseaux et Services de Mobiles) Mastére spécialisé program.

    Patrick Maillé: is graduated from Ecole polytechnique  and Telecom ParisTech in 2000 and 2002, respectively. He has been with IMT Atlantique since 2002, where he obtained his PhD in applied mathematics in 2005, and his habilitation (HDR, Rennes 1 university) in 2015. He has held visiting scholar appointments at  Columbia University  (June-December 2006) and UC Berkeley (academic year 2014–2015). His research interests are in all economic aspects of telecommunication networks, from pricing schemes at the user level, to auctions for spectrum and regulatory issues (net neutrality, search neutrality).

    View full text