IEEE SmartGridComm 22 Workshop - Cognitive and Machine-to-Machine Communications and Networking for Smart Grids Radio Resource Allocation for Group Paging Supporting Smart Meter Communications Chia-Hung Wei, Ray-Guang Cheng, and Firas M. Al-Taee Department of Electronic Engineering, National Taiwan University of Science and Technology Email: crg@ieee.org Abstract Smart meter is one of the novel applications in next generation cellular networks. Group paging can effectively solve the RAN overload control problem in 3GPP LTE, but the usage efficiency of the radio resource is needed to be improved. This paper proposed a dynamic radio resource allocation algorithm to solve the inefficiency problem. I. INTRODUCTION Smart meters are a physical technology that is added to or replaces a typical periodic meter. Most smart meters are computerized and allow for remote data collection through periodic communication to the utility on energy use. In this way, utilities can gather information on energy use on near real time basis [, 2]. A grid of smart meters along with other components form what is called a smart grid, a smart grid is a user-centric system which improves the flexibility, efficiency, accessibility and reliability of utilities. It uses two-way data communication technologies to provide the functions of data collection and providing consumers with data related to demand and fault report [3]. Therefore, the smart grid has attracted more and more attention in the last few years[3]-[8].in [3, 4] different wireless communication infrastructures used for smart grid are discussed. In [5] the authors had investigate the tradeoff of the cyber security issue and the power consumption for the remotely deployed sensors. The quality of experience (QoE)-driven power scheduling algorithm for smart grid has been proposed in [6]. By take the real environment into consideration, the scenario of smart meter density in central and urban areas of London have been provide in [7]. This scenario can be used in further investigation and evaluation. The traffic characteristics of smart meters have been studied in the literatures of [8]. It was found that the smart meters have high device density (e.g., up to per sector) and may periodically communicate with the central system in a synchronized manner [8]. Existing cellular network is one of the best candidates to carry the traffic generated by smart meters. However, the mass concurrent data generated by smart meters may result in the congestion at the radio access networks (RANs), which cause intolerable delays, packet loss or even service unavailability to existing human-tohuman (H2H) communication services [7]. Therefore, the network may require proper overload control mechanisms to guarantee network availability and quality of H2H services under heavy smart meter load [7]. The overload control of uplink random access channel (RACH) in radio access network (RAN) is one of the important working items for 3GPP Long Term Evolution (LTE) [7]. The purpose of RAN overload control is to avoid high random access collision probabilities when mass devices contend for RACH at the same time. Several push-based RAN overload control such as access class barring (ACB) scheme, and separate RACH resources for smart meter device, dynamic allocation of RACH resources, smart meter device specific backoff scheme, and slotted access have been proposed [7]. In ACB scheme, the network separates the smart meter traffic into several access classes and assigns an ACB factor to each smart meter access class. Each cell can control the channel access probability of the specific smart meter access class by setting the ACB factor. In the separate RACH resources for smart meter scheme, the network provides different access collision probabilities for the H2H and smart meter devices by providing dedicated RACH resources for them. In the dynamic allocation of RACH resources scheme, the network aims to dynamically allocate additional RACH resources for the smart meter devices based on the predicted access load of smart meter devices. The smart meter specific backoff scheme aims to delay the random access 978--4577-73-/2/$26. 22 IEEE 659
(re-)attempts of smart meter devices by assigning smart-meter-device-specific random backoff procedure. In slotted access scheme, each smart meter device is associated with dedicated access cycle/slots (similar to paging cycle/slots) through its identity (ID). Each smart meter device can perform the random access only at its access slot. The pros and cons of different push-based RAN overload control schemes were summarized in [9]. The pull-based RAN overload control [] is another possible solution. For pull-based RAN overload control, the smart meter server needs to activate a large amount of smart meter devices simultaneously. The base station (BS), which is also known as the enhanced Node B (enb) in 3GPP, or radio network controller (RNC) may use existing paging mechanism to page a large amount of smart meter devices. However, existing paging mechanism is designed for H2H service and may not be suitable for smart meter traffic. In LTE, a downlink paging channel is defined to transmit the paging information to a UE, inform UEs about a system information change, and Earthquake and Tsunami Warning service (ETWS) or Commercial Mobile Alert System (CMAS) notification. The network may transmit a paging message to activate a specific UE at UE's paging occasion. The paging occasion of each UE is determined based on its UE identity (UE-ID). In LTE, each paging message can only paging up to sixteen UEs and only two paging occasions are available in one radio frame []. Therefore, the network has to transmit multiple paging messages (i.e., due to the capacity limitation of the paging message) during a long period of time (due to the different paging occasions of the smart meter devices and the limited number of paging occasions in one radio frame) if the network wants to activate a large number of smart meter devices. Another approach of the pull-based RAN overload control is to use group paging [7]. In group paging, a BS may page a group of smart meter devices to access the network simultaneously. However, the simultaneous channel accesses from a large number of smart meter devices may result in access congestion in the random access channel. Hence, the network has to determine the proper group size and allocate enough radio resources to meet the performance constraint of the smart meter services. In this paper, we will only focus on the performance of group paging. Fig. System model of group paging for smart meter in LTE. Fig. 2 Flow chart of dynamic radio resource allocation algorithm Several studies of group paging have been presented in the literature. In [], the definitions of the collision probability used in analysis the performance of group paging had been classified. In [], a simple way to implement group paging is proposed. Details of the operation of group paging will be elaborated in Sec. II. A. In [2], a primary study of group paging via computer simulation is presented. It is shown that the group paging can effectively solve the RAN overload problem. In [3], a simple analytical model is proposed to dervie the performance metrics of a simplified version of group paging. However, the simplified group paging disables the random backoff mechanism and ignores the processing delay of base station as well as the transmission time of the message part. In all of the studies, the radio resource of the random access channel 66
(RACH) is static and will not be changed during the whole paging period. This paper presents a dynamic radio resource allocation (DRRA) algorithm of the group paging mechanism for LTE networks [2]. The performance metrics of collision probability, access success probability, defined in [7] and radio resource utilization were evaluated. The rest of this work is structured as follows. The system model of group paging is described in Section II. Section III shows the simulation results. Conclusions are finally drawn in Section IV. II. SYSTEM MODEL A. Group Paging This paper considers a group of M smart meter devices which are uniformly distributed in one cell, as illustrated in Fig.. It is assumed that the M smart meter devices are assigned by a group identity (GID) and the network utilize the GID to page the M smart meter devices to access the network simultaneously. The authors suggested that a smart meter device will be assigned by a unique group identity (GID) after camping on a network and joining the group. All of the smart meter devices in the group shall listen to the same paging channel at the same paging occasion derived from the GID []. The network may send a paging message containing the GID and a backoff time for the group of smart meter device to indicate the time of access from the reception of the paging message. Upon receiving a paging message containing the GID, the smart meter device shall follow the LTE standard random access procedure defined in [4] to establish a connection with the network. That is, the smart meter devices shall send a randomly chosen preamble at the time specified in the paging message. The enb will reply an random access response (RAR) message indicating result of the random access. The smart meter devices which are not acknowledged by the RAR message within random access response window are referred as collided smart meter devices [7]. All of the collided smart meter devices should perform random backoff and then re-transmit their random access attempts in the next random access slot. The process repeats until the maximum number of random access transmissions is reached. B. radio resource allocation As illustrated in Fig. 2, the enb assigns radio resources (preambles) on per random access slot basis. Pending devices are those devices which still have random access contention chance. It is defined as the group size excluding successful failing smart meter devices in a given time slot. In the above algorithm enb assigns a number of radio resources equal to the number of pending smart meter devices in each random access slot. C. Performance Metrics Several performance metrics were specified in [7] for evaluating the performance of the RAN overload control schemes. In this paper, the performance metrics of collision probability and access success probability were evaluated for group paging. The performance metrics are defined as [7]: - Collision probability: the ratio between the number of occurrences when two or more smart meter devices send a random access attempt using exactly the same preamble and the overall number of opportunities (with or without access attempts) in the period. - Access success probability: the probability to successfully complete the random access procedure within the maximum number of preamble transmissions. III. Simulation Results Computer simulations were conducted to evaluate the performance of group paging based on DRRA algorithm, and it was further compared with static radio resource allocation (SRRA) algorithm. The SRRA algorithm is to let the radio resource assigned by enb in each random access slots are same. We implemented the group paging protocol in our simulation using C language. The RACH parameters and the random access procedures confirm to the LTE standard specification [7, 4, 5]. The basic parameters for RACH capacity for LTE FDD specified in [7] were used in the simulation and are summarized in Table. In SRRA algorithm, it was assumed that a enb reserves a preamble set of fifty-four preambles per random access slot. In DRRA algorithm, the number of preamble reserved by the enb may dynamic adjust according to the number of pending smart meter. An error-free wireless channel without capture effect was 66
considered in the simulations. In Figs. 3 to 5, the results for the paging group size (M) of smart meter devices from fifty to one thousand were demonstrated. In the simulations, each point represented the average value of 7 samples. Each sample was obtained by performing group paging. The simulation was start from the time all the smart meter devices in the paging group simultaneously receive the paging message, and end at the time that all the smart meter devices are either success or fail. Fig. 3 shows the collision probability as a function of the paging group size of smart meter devices. The results show that the collision probability increases as the number of paged smart meter devices grow. It is also found that the collision probability for DRRA algorithm is much higher than SRRA algorithm in small paging group size. It is because DRRA algorithm reducing much more radio resources in the case of small paging group size. Fig. 4 shows the access success probability. The results show that the RACH configuration of Table can support paging group size up to 5 smart meter devices with an access success probability of.98. The results also show that the access success probability of.98 can be achieved even if the collision probability exceeds.3. In the following, we tried to find out the usage efficiency of the radio resources assigned by enb during the whole paging period by simulation. The purpose of DRRA algorithm is to prevent the enb provide too much radio resources when the pending smart meter devices decrease. In the other words, enb may save radio resources by providing proper amount of radio resources to the networks. For SRRA algorithm, the radio resources assigned by enb always keep the same, it will be wasted/unused when the pending devices going down. The usage efficiency of the radio resources were observed from other two performance metrics defined by us, they are total number of assigned radio resource and average utilization. The total number of reserved radio resource is defined as the total number of radio resources assigned by enb within whole paging period. As shown in Fig. 5 the enb can aggressively reduce the usage of radio resources when paging group size is not greater than six hundred. For those paging group size which are greater than six hundred, the total number of assigned radio resources are identical, it is because of the pending smart meter devices didn t decrease until the end of the paging period.. IV. CONCLUSIONS Group paging is one of the solution proposed to resolve the RAN overload control problem resulted from smart meter traffic. However the usage efficiency of the radio resources for group paging is quite low. This paper proposed a dynamic radio resource allocation algorithm in order to make the enb assign a proper amount of radio resources to the networks. We also presented the performance of dynamic radio resource allocation for group paging via computer simulations. The results showed that the enb can saved 4% of radio resource and keep the access success probability in % and increase radio resource utilization for 3% when activating 4 smart meter devices. The simulation result can be used for the enb as a reference to determine the group size and the allocated radio resources in order to meet the service constraints (e.g., access delay and access success probability) of the smart meter applications. Further simulations or analytical studies need to be conducted to explore the tradeoffs and design strategies of the radio resource allocation for each random access slot. Table : Basic simulation parameters for RACH in LTE FDD Parameter Setting Cell bandwidth 5 MHz PRACH Configuration Index 6 Total number of preambles 54 Maximum number of preamble transmission Ra-ResponseWindowSize 5 subframes mac-contentionresolutiontimer 48 subframes Backoff Indicator 2ms HARQ retransmission probability for Msg3 and Msg4 (non-adaptive HARQ) % Maximum number of HARQ TX for Msg3 and Msg4 (non-adaptive HARQ) Processing delay of HARQ ACK for Msg3 and Msg4 (unit: sub-frame) Gap of receiving Msg4 Gap of Msg3 retransmission (unit: subframe) 4 Gap of Msg4 retransmission (unit: subframe) 5 4 662
Collision probability Access success probability.9.8.7.6.5.4.3.2. Static 2 3 Fig. 3 Collision probability.9 Static.8.7.6.5.4.3.2. 2 3 Total number of assigned radio resources 3.5 x 4 Static 3 2.5 2.5.5 Fig. 4 Access success probability and Methodology, IEEE Communications Magazine, vol. 5, no. 5, pp. 36-4, May 22. [7] 3GPP TR 37.868, RAN Improvements for Machinetype Communications, v..7., Oct. 2. [8] 3GPP R2-234, Smart Grid Traffic Behaviour Discussion, Verizon, RAN2#69bis, April 2. [9] 3GPP R2-27, Evaluation on Push Based RAN Overload Control Schemes, Huawei and HiSilicon, RAN2#73bis, April 2. [] 3GPP R2-487, Pull Based RAN Overload Control, Huawei and China Unicom, RAN2#7, Aug. 2. [] R. G. Cheng, C. H. Wei, S. L. Tsao and F. C. Ren, RACH collision probability for machine-type Communications, IEEE 75th Vehicular Technology Conference, May 22. [2] 3GPP R2-398, Further Analysis of Group Paging for MTC, ITRI, RAN2#74, May 2. [3] C. H. Wei, R. G. Cheng, and S. L. Tsao, Modeling and estimation of one-shot random access for finite-user multichannel slotted ALOHA systems, IEEE Communications Letter, vol. 6, no. 8, pp. 96-99, Aug. 22. [4] 3GPP TS 36.32, Evolved Universal Terrestrial Radio Access (E-UTRA) Medium Access Control (MAC) Protocol Specification, v. 9.3., June 2. [5] 3GPP R-6584, E-UTRA Random Access, Ericsson, RAN#44, Feb. 26. 2 3 Fig. 5 Total number of reserved radio resources REFERENCES [] E. Doris and K. Peterson, Government Program Briefing: Smart Metering, NREL, US Dept. of Energy, Tech. Rep. TP-7A3-52788, Sep. 2. [2] D. Kathan et. al., Assessment of Demand Response and Advanced Metering, Federal Regulatory Commission, pp.4, September 2. [3] D. Niyato and P. Wang, Cooperative Transmission for Meter Data Collection in Smart Grid, IEEE Comm. Mag. vol. 5, no. 4, pp. 9-95, April 22. [4] C. R. Yu. Y. Zhang, S. Gjessing, Y. Chau, S. Xie, M. Guizani, "Cognitive Radio based Hierarchical Communications Infrastructure for Smart Grid", IEEE Network, vol. 25, no. 5, pp. 6-4, Sept./Oct. 2. [5] A. M. Qiu, H. Su, M. Chen, Balance of Security Strength and Energy for PMU Monitoring System in Smart Grid, IEEE Communications Magazine, vol. 5, no. 5, pp. 42-49, May 22. [6] B. L. Zhou, J. Rodrigues, and L. Oliveira, QoE-Driven Power Scheduling in Smart Grid: Architecture, Strategy, 663