Fairness-constrained optimized time-window controllers for secondary-users with primary-user reliability guarantees
Introduction
Vehicular networking is in the progress of merging with the Mobile Internet and Mobile Cloud Computing (MCC), so as to constitute an integrated communication/computing information platform [1]. Recently, several solutions were proposed to address the expected mission of the next-generation of vehicular networks [2], and Vehicular Cloud Computing (VCC) is emerging as the most appealing solution. About the spectral crowding of current vehicular networks, the communication traffic flows typically generated by safety applications and routed in downlink over the available backbones are of burst-type. These flows present inter-arrival gaps of non-negligible time durations, which could be opportunistically exploited by Vehicular Clients (VCs) equipped with Cognitive Radio (CR) [3] smartphones to access to the serving Road Side Units (RSUs) through WiFi connections [4]. However, in order to continue to guarantee higher priority to the traffic flows generated in downlink by the service providers, hard upper bounds must be imposed on the tolerated instantaneous rate of packets collision among the backbone traffic (that is, the primary traffic) and the access traffic (that is, the secondary traffic). In this paper, we focus on Cloud and Internet-assisted Vehicle-to-Infrastructure (V2I) communication for emerging non-safety applications where multiple car smartphones equipped with heterogeneous cognitive capabilities play the role of secondary-users, and compete for acceding to the serving RSUs by opportunistically exploiting the time and frequency holes of the traffic flow generated by the service provider. It is expected that these bandwidth-demanding applications will have a significant impact on the commercial success of VCC and will contribute to accelerate its implementation and deployment [1], [5]. This work is for some issues the continuation of the work [6], even if it is almost completely self-contained. In [6], we pursued the twofold target of the joint maximization of the steady-state memoryless aggregate access goodput of the overall network and the average per-client access rates, mainly focusing on the coupling among the two subproblems. In this paper, we mainly address the Medium Access Control (MAC) part of the vehicular problem, over a broader area of investigation and with a deeper analytical detail. We study different families of controllers and derive new theoretical results about their hierarchy of performances (throughput-gain, performance bounds, and not immediately intuitive situations where the more constrained controllers we propose do not present any optimality gap compared to more standard less reliable approaches). Due to space limitation, in the following we refer to [6] for a detailed explanation of the vehicular and channel model, (i.e., the application framework we considered in the numerical result section to test our controllers). Details on physical layer aspects, VC optimal management and the adaptive optimal control of the traffic flows and energy of the VCs can be found in [6], [7], [8]. In this paper, we only expose the new technical aspects and results.
From an optimization point of view, in this paper we focus on the RSU access time-window allocation problem. We do not tackle here in detail the way each VC autonomously computes own desired access rate. We address the MAC problem without handling the client local optimization of the requested throughput, here considered as a fixed (even if slot-by-slot time-varying) optimization input parameter. From a system point of view, the client request phase, the ACK/NACK phase and the other phases of a vehicular protocol are implemented in the simulation result scenario.
More in detail, in Section 3 we formalize the target of the RSU access time-window control and we define two different access time-window optimization problems, depending on the primary-user QoS requests. In Section 4, we derive the related optimal soft and hard steady-state controllers, and discuss the subclass of the memoryless controllers. In Section 5, we derive closed form expressions for the performance of schedulers, and for the access-goodput gain with respect to the memoryless controllers. We also prove as hard controllers do not have any optimality loss with respect to the soft one, and we discuss the implementation and complexity aspects, showing as the hard controllers are able to make the outage probability vanishing and adapt on-line to the occurring cluster changes, two major challenges for real-time applications.
Finally, in Sections 6 and 7 we extend the obtained results to fairness constrained problems (and problems able to take into account for service level differentiation among VCs), then we integrate our controllers with cognitive data-fusion techniques. Finally, in Section 8, we present numerical results and controller comparisons.
Section snippets
Related work
In recent years, there have been burgeoning research efforts involving cognitive radio networks. We review only a few of them, focusing on the most related works. In [9], the authors presented a new energy efficient hybrid MAC for wireless sensor networks with QoS guarantees. Recently, the authors in [10] proposed a new approach which aims to design MAC protocol for enabling Secondary Users (SU) to opportunistically access the idle time slots in a primary TDMA network. In this paper, two
System model and soft/hard time-window optimization problems
In the following, time is slotted, Ts (s) is the slot duration, t is the discrete-time slot index. The VCs may change their spatial positions and the network nodes may start to transmit only at the beginning of each slot t, and i is the cluster index. In order to cope with the aforementioned bandwidth limitation, the communication traffic over the wireless backbone is Time Division Duplexed (TDD) and it is organized into super-frames of duration TSF (s). Each super-frame comprises an uplink
The optimal steady-state controllers
After introducing the dummy positions: let the secondary-query event, that is, the event that at-least one client requires to transmit on the slot t in cluster(i), where is the corresponding steady-state secondary-user request probability. Let be the secondary-access event, that is, the event that the time-window controller decides to allow at-least one secondary-user to access to the channel in slot t, with being
Access-goodput gain and complexity aspects
By comparing now the goodput provided by the memoryless controller (11) with the general one of Proposition 3, for the goodput gain: we obtain:showing as the gain of the general controller with respect to the memoryless one is strictly related to the target secondary-user access probability in (13). In a simple case, when all the VC throughput requests are uniformly distributed between 0 and a maximum value rmax , (16) can
Maximum system throughput in presence of SLD/fairness probabilistic constraints
We defined the objective function in (2) so to take into account for the client weight vector and we called it utility or goodput of the system. Based on the secondary-user QoS requirements, mobility, traffic and channel conditions, the VC-weights θj can be properly chosen so to assure fair access and/or service level differentiation (SLD) among clients. The vector θ could also be a priori fixed based on several economical aspects and/or differentiated service among clients.
Primary user activity estimation and time-variant time-window controllers
In this section, we integrate the optimal steady-state time window controllers we derived in previous section with the primary user activity estimation techniques implemented by cognitive devices, and we discuss why this is a time-variant scenario with possibly no steady-state, so that it demands for new optimized solutions.
Simulated scenario and performance tests
In the first part of our carried out tests we consider a homogeneous networked vehicular infrastructure, where all RSUs and VCs exhibit the same performance capabilities and mobility features. As simulation values, when not otherwise specified, we considered (ms), a number of clusters equal to and an average number of client per-cluster per-slot equal to . As in [28, Section 3.2], we resort to the so called Markovian random walk with random positioning for
Conclusion
We developed optimal controllers for the adaptive and scalable management of the bandwidth resources in vehicular access networks that provide reliability guarantees to the primary-user. We optimized classes of controllers able to handle the outage probability without any loss in performances, and obtained closed-form expressions for the aggregate overall throughput and throughput-gain. We generalized results to fairness constrained problems and cognitive scenarios, showing as controllers are
Acknowledgement
This work has been supported by the project no. 2015YPXH4W_004: “GAUChO - A Green Adaptive Fog Computing and Networking Architecture” funded by Progetti di Ricerca di Rilevante Interesse Nazionale (PRIN) Bando 2015, and by the projects: “Vehicular Fog for energy-efficient QoS mining and dissemination of multimedia Big Data streams (V-FOG and V_FOG2)”, funded by Sapienza University of Rome.
References (28)
- et al.
A survey on vehicular cloud computing
J. Netw. Comput. Appl.
(2014) - et al.
Stochastic traffic engineering for real-time applications over wireless networks
J. Netw. Comput. Appl.
(2012) - et al.
Energy efficient and QOS aware medium access control for wireless sensor networks
Concurr. Comput. Pract. Exp.
(2010) - et al.
Vehicular networking: a survey and tutorial on requirements, architectures, challenges, standards and solutions
IEEE Commun. Surv. Tutor.
(2011) - et al.
Advances in cognitive radio networks: a survey
IEEE J. Sel. Top. Signal Process.
(2011) - et al.
Now–network on wheels: project objectives, technology and achievements
Proceedings of International WIT
(2008) Cognitive radios and the vehicular cloud
Proceedings of Workshop on Cognitive Radio Architectures for Broadband
(2013)- et al.
Reliable adaptive resource management for cognitive cloud vehicular networks
IEEE Trans. Vehic. Technol.
(2015) - et al.
Distributed and adaptive resource management in cloud-assisted cognitive radio vehicular networks with hard reliability guarantees
Veh. Commun.
(2015) - et al.
Cognitive medium access control protocols for secondary users sharing a common channel with time division multiple access primary users
Wireless Commun. Mob. Comput.
(2014)
Primary service outage and secondary service performance in cognitive radio networks
Wireless Commun. Mob. Comput.
A MAC sensing protocol design for data transmission with more protection to primary users
IEEE Trans. Mob. Comput.
Decentralized cognitive MAC for opportunistic spectrum access in ad hoc networks: a POMDP framework
IEEE J. Sel. Areas Commun.
Cross-layer based opportunistic MAC protocols for QOS provisionings over cognitive radio wireless networks
IEEE J. Sel. Areas Commun.
Cited by (3)
Optimisation of computer network reliability based upon sensor technology and genetic algorithm
2023, International Journal of Global Energy IssuesAdvanced and Innovative Optimization Techniques in Controllers: AComprehensive Review
2021, International Journal on Recent and Innovation Trends in Computing and Communication