Improved Method of Contention-Based Random Access in LTE System

Han-Song Su, Min Zhang, and Gao-Hua Liu

Abstract—Random access is the necessary process to establish the wireless link between the user equipment(UE) and network. The performance of the random access directly affects the performance of the network.In this work, we propose a method on the basis of the existing alternatives. In this method, we estimate the system load in advance to adjust the number of terminals. An access threshold is set to control the number of terminals which want to access the base station at an acceptable level. At the same time, we have an improvement on the existing power climbing strategy. We suppose that the power ramping is not always necessary for the re-access. And the selection of power ramping steps is studied in this paper.Simulations based on MATLAB are employed to evaluate the effectiveness of the proposed solution and to make comparisons with existing alternatives.

1. Introduction

The third-generation partnership project (3GPP) long-term evolution (LTE) is a high data rate, low delay, and packetbased mobile communications system. One of its key technologies is the random access procedure. The procedure starts with the transmitting of random access preamble, which is generated by the cyclic shift of the Zadoff-Chu (ZC)sequence for user identification, from the user equipment (UE)to the base station[1]. Compared with the non-contention-based random access, in the LTE system, the more commonly used method is the contention-based random access, in which a plurality of UE can try to access the channel simultaneously and the preamble collision is likely to happen. The preamble collision is a very important reason for the performance degradation of communications systems, which directly leads to the reduction of the system throughput, access time delay,and the success rate of access. In particular, the increase of the UE will increase the probability of collision and the performance of the system will reduce dramatically[2].

In order to solve this problem, it is generally commenced from two aspects: The first one is to avoid the collision fundamentally; the second one is taking effective measures,such as some backoff policy, to minimize the impact of collisions. References [3]-[8] are addressed this problem from the first aspect avoiding the collision fundamentally. In the method illustrated in [3], the UE will not start a random access process for every single packet. Instead, it triggers a random access process when the number of aggregated packets reaches a given threshold. Reference [4] controlls the number of UE’s preamble transmission according to the load of the network.The basic idea is to limit the number of transmissions and retransmissions in order to reduce the collisions rate while keeping the stability of the system[4]. Reference [5] proposed a method in which the base station can measure the channel condition of “busy” or “idle” and broadcast them to UE and UE will start a random access process only when the channel condition is idle, which is also a good choice for reducing the collision probability. In this case, the random access preamble collision can be prevented effectively.These schemes are traditional and addressed in the medium access control (MAC) layer only, in [6], the authors combined the MAC layer transmission scheduling protocol with the physical layer signal-separation algorithm to resolve contentions. An adaptive pre-backoff proposed in [7] is also effective to reduce the collision rate.

For another aspect, when the preamble collision has happened, effective measures should be taken to reduce the impact. In the LTE system, ALOHA is the most popular random access protocol, with which UE retransmits a message after a certain time delay on the basis of a specified backoff policy if it failed in the access process[2]. And the slotted-ALOHA (S-ALOHA) protocol is the first choice under the condition of any delay and energy constraints[8]. In [2], the authors proposed a highly efficient backoff policy, namely“acknowledgment (ACK)-based adaptive backoff (AAB)”.This method takes into account the successful transmission numbers in neighboring stations when it chooses the backoff time before the next access. By doing this, ABB can effectively reduce the packet collisions in heavy-loaded conditions. Reference [9] determined the backoff time according to the network condition which is estimated by each UE using the model GM(1, 1) in the grey system. The improvement of the control algorithm which was needed in the S-ALOHA protocol was proposed in [10].

Additionally, the generalized likelihood ratio test criterion(GLRT) is also used to improve the random access algorithm[11]. The authors applied GLRT to decide whether a given code is presented or not in the physical random access channel (PRACH). Random access for machine-to-machine(M2M) communications is one of the current research hotspots.References [12] to [15] studied on the M2M system and some improvements of the random access procedure for the M2M system are presented, respectively. In [16], a reinforcement learning-based evolved node B (eNB, also known as the name of the base station in the LTE system) selection algorithm that allowed the MTC devices to choose eNBs was proposed to avoid congestion caused by random channel access of MTC devices.

In this paper, we proposed an improved random access method based on the access threshold. In this method, the base station (eNB) will estimate the number of UE before starting the random access process and the access threshold is set according to the number of UE. Each UE will estimate whether it can start the random access processing or have to wait for another chance by generating a random number. The number of UE can be controlled in a certain range by this method so it can reduce the preamble collision fundamentally. At the same time, a study of the power ramping step is conducted in this paper. We suppose that the power ramping is necessary only when the random access processes failed, because the random access response (RAR) message receiving failure will trig a power ramping step. And we will propose a power ramping step size selection algorithm which outperforms the existing one.

The rest of this paper is organized as follows. Section 2 introduces the random access process briefly. Our improved random access algorithm will be proposed in Section 3 in detail and the power ramping method will be shown in Section 4. The performance of the proposed method is analyzed in Section 5 and the computer simulation result is shown in this part as well.Then, we will conclude our paper in Section 6.

2. Overview of Random Access

In this part, we consider the contention-based random access process which presents a promising solution for efficient uplink channel access. The random access which is produced between UE and eNB mainly consists of four steps. The detailed process was shown in Fig. 1.

2.1 Msg1

The process begins with the transmission of the preamble sequence (Msg1) which is generated by the circular shift of the ZC sequence from UE to eNB through PRACH. Before the transmission of Msg1, the messages, in which include the preamble index, PRACH preamble resource, radio network temporary identity (RA-RNTI), and preamble received power,will be broadcasted by eNB in system information block messages[16].

Fig. 1. Contention-based random access process.

2.2 Msg2

After the preamble is sent to eNB, UE will wait for the RAR message which must be sent within the response window[15]. Msg2 may contain various preambles transmitted from different UE, and can respond a number of UE’s random access requests simultaneously. The RAR message is considered receiving successfully only when the consistent preamble identifier sent in Msg1 is detected in the RAR message. If the UE failed to receive the RAR message, it means that the collision occurred. In this case,the next access procedure is performed after a time delay which is selected between 0 and backoff time. When the number of random access is to the limit times, the random access procedure is recognized failed and the access problems will be reported to the radio resource control(RRC) layer.

2.3 Msg3

In the case of UE receive the RAR message successfully,UE gets the resource that is distributed by eNB for Msg3 and send the Msg3 to eNB on PUSCH. The content of Msg3 is different in different access conditions. For the initial access,Msg3 includes a RRC connection request message. The hybrid automatic repeat request (HARQ) procedure is supported if the Msg3 transmission is failed.

2.4 Msg4

The random access procedure is completed when UE receiving a contention resolution message (Msg4) from eNB.This message is carried on the physical downlink shared channel (PDSCH). When the Msg4 transmission is failed, the HARQ procedure can be used for retransmission. The UE will feedback ACK only when the Msg4 is successfully received.

3. Improved Random Access Algorithm

In the LTE random access process, the increase of the UE will increase the probability of collisions and the performance of the system will be seriously affected. Let part of the users wait for a period of time before access can control the network load in a reasonable range. In this part, we propose an improved random access method based on this viewpoint. The realization of this method is mainly discussed in two aspects:The base station side (that is the eNB side) and the UE side.

3.1 Base Station Side

The main work of eNB is to estimate the number of UE and then set the access threshold according to the UE number.The ZC sequence is used as the preamble for random access procedure so that preamble could be detected efficiently and clearly because of the excellent auto-correlation and crosscorrelation characteristics of the ZC sequence[1]. In this paper,the frequency-domain detection is adopted when eNB receives the preamble signal to get the preamble sequence and transmitting power. The basic principle of the frequencydomain detection algorithm is shown as Fig. 2[17]. The signals received by eNB is

Fig. 2. Preamble detection principles.

where (·) means the point multiplication. Point multiply the complex conjugate of the received signalwith the local preamble sequenceand inverse discrete Fourier transform/inverse fast Fourier transform (IDFT/IFFT) with the result, then

Fig. 3. Relationship between the number of UE and mean of P/Pe.

Then, eNB can set the access threshold according to the number of UE. If few UE wants to access eNB, the eNB can set a low access threshold to enable more UE access the eNB.On the other hand, if the number of UE intending to access eNB is large, the eNB will set a high access threshold to control the number of UE accessing in a considerable range so as to reduce network congestion.

The selection of the access threshold is very important. If the access threshold is set too high, there is a little UE can start a random access process at the same time and most of UE has to wait for another chance to access the network, which will lead to the increase of the average time delay. While a low access threshold will hardly achieve a high success rate. That means the selection of access threshold should take into account both the success rate and time delay, instead of simply considering that the system will perform better with a high access threshold. The access threshold shown in Table 1 has been determined based on an overall consideration of the access delay and access success probability according to multiple simulation results. Taking the UE number equal to 80 as an example, Fig. 4 shows the success probability, collision probability, and average time delay in milliseconds (ms) at different access thresholds from 0.1 to 0.9. From the graph, we can see that when the access threshold is 0.3, the collision probability is the lowest, and when the threshold is 0.4, the success rate reaches the highest value and the average time delay reaches the minimum. So the access threshold is selected as 0.4 when the UE number is 80. The access threshold shown in Table 1 is aimed at the situations where the number of UEs is not more than 200 in general. The other access threshold value can also be set when the maximum number of UE is changed.

Table 1: System load and access threshold setting

Fig. 4. Performance with different access threshold (UE=80).

The access threshold will be filled into the system broadcast message and broadcast to each UE with PRACH configuration messages.

3.2 UE Side

UE will receive the broadcast messages which include the random access threshold before the UE sending preambles to eNB. Then each UE will generate a random numberxwhich is uniformly distributed between 0 and 1 and comparexwith the access threshold received with the broadcast messages. The UE can initiate a random access process only whenxis larger than the access threshold. Ifxis lower than the access threshold, the UE has to wait for another random access opportunity to initial a random access process again. Setting different access threshold according to the load of a network can keep the number of UE in a certain range of number, so that avoid the RACH overload to a certain extent. Having a pre-judgment of the random access channel load condition can reduce the probability of the preamble collision, thereby enhancing the success rate of UE random access.

The improved access process of the whole system is shown in Fig. 5. And the performance evaluation and the simulation result will be discussed in Section 4.

Fig. 5. Random access processing for UE side.

4. Proposed Method of Power Ramping

As mentioned in the introduction, in the random access process, there are two kinds of reasons may lead to the access process failure: The first one is UE failed to receive the RAR message which may mainly because of the low transmitting power; the second one is the contention resolution failure which is mainly because of a lot of UE that lead to the collision. In the 3GPP LTE system, when the random access fails, PRACH will ramp up the preamble transmit power to improve the success rate, that is, the preamble transmit power ramps up when the preamble is retransmitted. The calculation formula is given as

Table 2: Offset based on preamble format

But why UE fails to access eNB is not only because of the low transmit power, but also about the high probability of the collision. So improving the preamble transmitting power in this way may cause the waste of power resources and even the power interference, which will affect the quality of service(QoS) and the system performance.

For this problem, we suppose that only when the UE fails to receive the RAR message, the preamble transmit power ramps up at the next time when they start the random access process again, while those who have failed because of the collision will not need to ramp up the transmitting power. In Section 5, our proposed power ramping method is compared with the original method in the random access success probability, average time delay, and collision rate.

The choice of the power ramping step will affect the random access success probability of UE. A big power ramping step will increase the throughput of the cell, but it will result in the waste of power. So we propose a power ramping step selection algorithm. In this algorithm, the smallest power ramping step that can bring the highest success rate of random access process will be selected. The specific method will be described below.

In the LTE system, there are 64 preambles for UE to start a random access process. When the number of UE is less than 64, the probability of collision is very little so we can consider that the reason for the failure of random access, in this case, is the low transmit power. In our method, we suppose that when the UE number is less than 64 and we can ignore the collision between UEs, then the power ramping step that can bring the highest success rate of random access will be selected as the power ramping step of the system.

The simulation is employed in order to get the power ramping step. In this simulation, the number of UE is 60, and we calculate the failed times of random access, when the power ramping step is. The noise is modeled as the white Gaussian noise. Multiple simulation results show that the power ramping step will be different with the varying of noise power. As we can see in Fig. 6, When the noise power is 5 dBW, a power ramping step=0.5 dB can make the random access fail times as 0. So we select 0.5 dB as the power ramping step in case of noise power is 5 dBW. When the white Gaussian noise is 7 dBW, the power ramping step=0.6 dB can make the random access fail times as 0. In order to facilitate the calculation, we select 1 dB as the power ramping step in the case of noise power is 7 dBW. More about the selection of power ramping step is shown in Table 3.

5. Performance and Simulation Result

5.1 Performance Evaluation

The improved random access method proposed in this paper can reduce the access collision rate and increase the success rate of random access without changing the 3GPP LTE random access procedure. And the power control method prevents the power ramping leaded by the failure of resolving collisions, so that avoid the waste of power and the power interference generated by the power ramping process. Estimate the number of UE in advance to reduce the message transmit expense caused by the collision of a lot of UE. MATLAB simulations are employed to verify the performance of the proposed method. Mainly simulation output parameters are the average delay of random access, random access success probability, and random access collision probability. The formula of the random access average time-delay and success probability are

5.2 Simulation Conditions

The random access preamble format 0 and additive white gaussian noise (AWGN) channel model is used in the simulation. The simulation scene is the initial access, and the frame structure is the TDD structure. More simulation parameters are shown as Table 4.

Fig. 6. Selection of power ramping step with different noise power: (a) N=5 dBW, (b) N=7 dBW, (c) N=9 dBW, and (d) N=11 dBW.

Table 3: Selection of power ramping step

Table 4: Simulation conditions

5.3 Simulation Results

Random Access Threshold

In this simulation, we compared the improved method which is with the access threshold with the existing LTE random access method which is without the access threshold.The access threshold is set as Table 1 and the power ramping step is 1.

As shown in Fig. 7 (a), we can see that in the original method, the LTE random access time delay will increase with the number of users, and may eventually lead users cannot access the network for a long period of time, resulting in the network congestion. The proposed method performs a similar result when the user number is fewer than 60, while it outperforms the original method when the number of UE is increasing. It can control the time delay in an acceptable range with the rising number of users, the performance has been greatly improved.

As shown in Figs. 7 (b) and (c), we can see that in the original method of LTE, the collision probability is increasing obviously with the number of users constantly increased, as a result, the random access success probability will continuously reduce and ultimately it will be difficult for users to access the network. The proposed method can effectively reduce the collision probability especially when the user number is larger than 60, and can still maintain a high success probability with the increasing number of users.

Power Ramping

Fig. 7. Comparison between two methods with and without access threshold: (a) average time delay, (b) success probability, and (c)collision probability.

In this simulation, the original power ramping method is compared with the proposed method in the time delay, collision probability, and success rate of random access. And in the original method, the transmitting power will climb up every time when the random access preamble is retransmitted. In our proposed method, the power ramping is only taken in the case that UE fails to receive the RAR message and try to start a random access again. The power ramping step is set as 1. The result shows as Fig. 8.

Fig. 8. Comparison between original method and improved method with power ramping: (a) average time delay, (b) Success probability, and (c) collision probability.

As we can see in Fig. 8, the results show that the performance of these two methods are basically the same, that means our method can achieve the equal performance in terms of the time delay, collision probability, and success probability as the original one, whereas our method is energy efficient,since the power climbing generated by the collision is avoided.

Power Ramping Step

Then the performance of random access with power ramping steps respectively equal to 0.5 dB and 1 dB was simulated. In this simulation the white Gaussian noise is 3 dBW. The power ramping strategy is the original method, that is, the power ramps up every time when the random access preamble is retransmitted. The result shows as Fig. 9.

Fig. 9. Comparison between power ramping step=0.5 and 1: (a)average time delay, (b) success probability, and (c) collision probability.

The results show that the performance of the random access with the power ramping step of 0.5 dB equals to that with the ramping step of 1 dB. This confirms that choosing a suitable power ramping step can maintain the performance of the system at a considerable level with less power resource.

In a word, choosing a suitable power ramping step and ramping the power only when the random access process is failed because the user have not received the RAR message can save the power cost without reducing the system performance.

6. Conclusions

In this paper, the improved random access algorithm for the LTE system works with estimating the system load in advance and setting the access threshold according to the network load.In this way, we can to lead the network to be in a light load condition and avoid the network congestion which will affect the transmission of data or service. Moreover, we propose a method that can avoid the power ramping when the random access failed because of the collision of UEs. In the traditional method, the system will increase the transmitting power in this situation. The MATLAB simulations are employed to evaluate the effectiveness of the proposed solution and make comparison with the existing methods at the same time. The numerical simulations results show that the proposed random access algorithm that sets an access threshold in advance can effectively reduce the average time delay and improve the success probability of random access on some degrees, and the proposed power ramping strategy can save the power resource without affecting the performance of the system and QoS.