1. Introduction
With the advance of Intelligent Transport Systems (ITS), Vehicle to Everything (V2X) communication has been attracting attention. V2X can be roughly divided into wireless LAN-based Dedicated Short Range Communication (DSRC) and cellular-based Cellular V2X (C-V2X). The former has been standardized as IEEE 802.11p [
1] in the United States in 2010 by extending the IEEE 802.11 standard [
2]. In Japan, ARIB STD-T109 [
3] has been standardized by the Association of Radio Industries and Businesses (ARIB) in 2012 after formulating the experimental guidelines, ITS FORUM RC-006 [
4]. On the other hand, C-V2X has been standardized as Long Term Evolution (LTE) V2X based on existing LTE by Third Generation Partnership (3GPP) release 14 in 2017. LTE V2X supports the uplink/downlink communication which corresponds to Vehicle to Network (V2N) and the sidelink communication, which corresponds to V2V, V2I, and V2P, as shown in
Figure 1. Ref. [
5] analyzes scenarios of the C-V2X and gives a survey of the related articles. Modes 3 and 4 are available as ways to use radio resources in the sidelink communication [
6,
7]. In Mode 3, an evolved Node B (eNB) allocates the resources to User Equipment (UE) under network coverage. In Mode 4, the UEs autonomously select the resources without the need for control by the eNB.
The UEs communicate with each other by sensing-based Semi-Persistent Scheduling (SPS) for LTE V2X Mode 4. For autonomous resource selection, packet collisions are reduced by selecting resources that are not expected to be used by other UEs from reservation information and a measurement of the received power. However, because the resource selection depends on past information, the resource newly selected by one UE cannot be detected by other UEs until transmission on the resource. Therefore, the packet collision may occur because the UEs that select a resource at the same time cannot estimate the resources selected by each other. Regarding an evaluation of the received power, the resource released at the reselection cannot be effectively utilized because the evaluated value is smoothed without information about the resource reselection of other UEs.
Basic performance evaluation of Mode 4 is verified in a highway scenario [
8,
9,
10,
11] and a urban scenario [
12,
13]. The evaluation of congestion control and the impact of parameters are presented in [
14,
15] and [
16,
17,
18]. In addition to these works, analytical models are provided in [
19,
20]. On the other hand, methods to reduce the packet collisions are proposed in [
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31]. In [
21,
22,
23], each of the UEs notifies other UEs of resources virtually reserved by the UE in advance. In [
24], the UEs adjust the timing of reselection with other UEs. In [
25], the hidden terminal problem is mitigated by sending additional acknowledgments and feeding this information back to the transmitter. As a similar approach to Mode 3, a cluster-based resource allocation has been proposed in [
26]. However, these methods are not preferable from the viewpoint of compatibility with the standard because they require additional information and the format of control information needs to be changed. Geo-based resource allocation methods corresponding to the position of the UEs on a road have been proposed in [
27,
28,
29], and can avoid packet collisions with the UEs located on Non-Line of Sight (NLOS). Even so, it is necessary to know a correspondence between the UE position and the resource pool in advance, and it is difficult to achieve a high degree of efficiency unless all UEs follow the same rule. In [
30], although the UEs use alternately two different resources to reduce consecutive packet collisions, it is difficult to estimate the resources used by other UEs. A nonlinear averaging method has been proposed in [
31], which assigns higher priority to the latest information in evaluating the received power to effectively utilize the resources released by other UEs. Nevertheless, the effect is not sufficient for the reselection that occurred immediately before evaluating the received power. An enhancement to the SPS algorithm has been introduced in which each vehicle announces the reservation information for the reselection in [
32]. A balanced resource allocation scheme has been proposed in [
33].
In this paper, we propose two resource reselection methods in consideration of other UE reselection, using only the information originally included in the control information [
34]. In the proposed methods, the UEs can use the released resources by evaluating the received power only for the resources after other UEs’ reselection. Furthermore, it can avoid packet collisions by selecting from the resource range not selected by other UEs. The main novelty of this paper is the effective use of information contained in the already operational Resource Reservation Interval (RRI) in the wireless resource allocation algorithm using the SPS. Various methods have been studied to improve the performance of the wireless resource allocation problem by using additional information, but these methods require extensions that do not conform to the standard. In contrast, this study proposes and examines a new method for using the value of RRI included in the standard to improve performance while maintaining compatibility with the standard. The proposed method is highly feasible because it does not require any special extensions to existing standards. Since our method is designed to maintain compatibility with existing standards, it may or may not have a significant effect, but there are no disadvantages.
2. LTE V2X Mode 4
In the sidelink communication, the UEs transmit a packet in synchronization with each other by using the Global Navigation Satellite System (GNSS), eNB, and UEs [
35]. The sidelink communication supports a repeated transmission up to two times using Hybrid Automatic Repeat Request (HARQ) as an option. For Mode 4, which does not require the eNB control, it is optimized to autonomously avoid the packet collisions for periodic transmissions.
2.1. Radio Resources
Figure 2 shows an example of the resource usage for sidelink communication. A radio resource consists of a subframe (1 ms) and subchannels divided between the time domain and frequency domain. A subchannel consists of multiple Resource Blocks (RBs) of 180 kHz. The figure shows an example when the number of the subchannel is 2, and each UE transmits using the resource represented by the rectangle. The UEs transmit control information and data using adjacent frequencies. The control information is transmitted as Sidelink Control Information (SCI) [
36] over Physical Sidelink Control Channel (PSCCH). Although the SCI can notice periodic resource reservations, it cannot be used to reserve reselected resources. The data are transmitted in Transport Blocks (TBs) over Physical Sidelink Shared Channel (PSSCH). The TB is assumed to contain Cooperative Awareness Message (CAM) [
37], Decentralized Environmental Notification Message (DENM) [
38], and so on.
The structure of SCI Format 1 used in Mode 3 and 4 is illustrated in
Table 1. SCI Format 1 includes the following information:
Priority: the priority of the packet set by upper layers;
Resource Reservation: the time period of the intended next use of the resource;
Frequency resource location of initial transmission and retransmission: Resource Indication Value (RIV) corresponding to a starting subchannel index and length in terms of contiguously allocated subchannels;
Time gap between initial transmission and retransmission: subframe differences in repeated transmissions;
Modulation and coding scheme: the combination of modulation scheme and code rate;
Retransmission index: a boolean denoting whether the first transmission or second transmission.
The resource reservation value is interpreted for the other UEs as
X according to
Table 2, where it is equal to RRI divided by 100. The RRI represents the transmission cycle, which is basically 100 ms, and 20, 50, 100, 200, …, 1000 ms can be configured [
6].
In the sidelink communication, the UEs transmit by Single Carrier Frequency Division Multiple Access (SC-FDMA) like LTE uplink. A subframe consists of 14 symbols that are time-multiplexed as shown in
Figure 3 and
Figure 4, where the last symbol is used for Tx/Rx switching and timing adjustment [
39]. Two DeModulation Reference Signals (DM-RS) are contained in a subframe in Release 12 as shown in
Figure 3. To improve the demodulation accuracy in a fast-moving environment, the DM-RS is increased to four symbols in Release 14 as shown in
Figure 4.
2.2. Sensing-Based SPS
In Mode 4, the UEs select a resource using the sensing-based SPS. The SPS is performed by sensing for the past one second, and resource reservation information is notified by PSCCH. The UEs set Reselection Counter (RC) depending on the RRI from a uniform distribution in the range of for , for , and for . The RC is decremented every transmission. The UEs use the same subchannel resources until the RC reaches 0. In addition, when it is equal to 0, they decide whether to reselect the resource based on the probability of keeping their resource where can be set in the range . If the UE decides to reselect the resource, a selection window is set in the time interval for subframe n where and . A candidate resource in selection window is defined with subchannel x and subframe y where and . Then, the UE initializes the set with all candidate resources and excludes them according to the following steps a, b, and c.
In Step a, the UE excludes unmonitored resources due to the constraint of half-duplex transmission. In Step b, the UE excludes any resources reserved by other UEs. When the packets are received with PSCCH Reference Signal Receive Power (PSSCH-RSRP) of the TB corresponding to the SCI greater than or equal to threshold
and satisfy
where parameter
m is a subframe in sensing window defined by
and
is the RRI notified over PSCCH, they represent the reserved resources. When the resource is reselected,
is set to 0 because the UE does not make a reservation.
is set by upper layers depending on priority
of the transmitter and
of the receiver (both within 0 and 7). It can be in the range
dB and can be expressed as [
6,
40]
After the exclusion by Steps a and b, the UE checks whether the remainder of
contains at least 20% of the total number of candidate resources. If this condition is not satisfied, then Step b is repeated with
increased by 3 dB. In Step c, the UE excludes any resources whose received power is highly evaluated. Metric
is evaluated as the linear average of Received Signal Strength Indicator (RSSI) that for
can be expressed as
The UE moves with the smallest from to the set initialized to an empty set. This step is repeated until the number of candidate resources in reaches greater than or equal to 20 % of the total number of candidate resources. After these steps, the UE randomly selects one of the resources in .
An example of the exclusion process by Steps a, b, and c is demonstrated in
Figure 5. In this figure, the UE excludes the candidate resources in subframe 1020 by Step a because the UE has transmitted in subframe 420. In addition, the UE excludes it in subframe 1050 and subchannel 0 because the UE receives a packet with
in subframe 950 and subchannel 0. Similarly, the UE excludes it in subframe 1080 and subchannel 1 because the UE largely evaluates its RSSI.
2.3. Evaluating RSSI for Released Resources
In Step c, the UE evaluates the RSSI by the linear average in units of 100 subframes up to 1000 subframes in the past. If other UEs reselect a resource, the RSSI for the released resources may be greatly evaluated since the average value includes the resources that have been used before their reselection. In such a case, they are no longer candidates for the UE despite the resources released by other UEs. As a result, it results in inefficient use of resources.
Figure 6 shows an example of the RSSI evaluation for the released resources. This figure indicates that UE2 has reselected in subframe 820 and UE1 reselects in subframe 1000. Now,
is a candidate resource for UE1 corresponding to the resource released by UE2 in subframe 820. When evaluating the RSSI for
, UE1 largely evaluates it because it includes the resources before UE2’s reselection.
may not be able to be included in
since
is composed of resources evaluated to be small RSSI.
2.4. Overlapping Selection Window
In a resource selection by the SPS, the UEs select a resource to avoid packet collisions with each other. The resources used by other UEs are estimated based on a feature of using the same subchannel resources periodically. Hence, if a UE reselects a resource, other UEs cannot detect which resource was selected until the transmission using the newly selected resource is performed.
Figure 7 shows an example of an overlapping selection window. This figure indicates that UE2 has reselected in subframe 950 and UE1 reselects in subframe 1000. The range of the selection window for UE2 is from a maximum of subframe 951 to 1050, and that for UE1 is from a maximum of subframe 1001 to 1100. In this case, the section between subframe 1001 and 1050 is overlapped, and UE1 may select the same resource as UE2.
3. Proposed Method
In this paper, we propose new two resource selection methods in consideration of the fact that notified over PSCCH reaches 0 when the UEs reselect. The proposed methods can reduce the packet collisions without constraints such as additional information.
3.1. Partial Averaging (PA) Method
We propose a Partial Averaging (PA) method to address the problem of large RSSI evaluations for the released resources. The PA method evaluates the RSSI only for resources after receiving a packet with . This method enables the UEs to prevent the RSSI from being greatly evaluated by limiting the smoothing range to the resources and be able to make effective use of the released resources.
Figure 8 shows an example of the RSSI evaluation using the PA method. This figure indicates that UE2 has reselected in subframe 820 and UE1 reselects in subframe 1000. When evaluating the RSSI for
in the SPS, UE1 largely evaluates it because of averaging resources before subframe 820 that was used by UE2. In contrast, the PA method evaluates the RSSI only for resources after subframe 920 since the packet with
has been received in subframe 820. In subframe 920, UE2 has not transmitted, and the RSSI for
is evaluated to be smaller. As a result, UE1 can make effective use of
.
3.2. Overlapping Avoidance (OA) Method
In this section, we propose an Overlapping Avoidance (OA) method to address the problem of the overlapping selection window. In the OA method, the UE first defines the subframe as in which the last packet with was received within the last 100 subframes from the resource selection. Then, the UE moves , which is a subset of , to if and if , respectively. Finally, the UE randomly selects a resource from if and if . Now, and mean the set of resources that can be selected by the UE that transmits the packet with and the set of resources in which the UE has free from the possibility of the selection, respectively. Note that the selection window of the UE is up to 100 subframes, but the resources with are included in because the resources in are excluded by Step a.
Figure 9 shows an example using the OA method. This figure indicates that UE2 has reselected in subframe 950, and UE1 reselects in subframe 1000. In this case,
because UE1 has received a packet with
in subframe 950. UE1 divides
into
and
, where
includes the candidate resources between subframe 1001 and 1049, and
includes the candidate resources between subframe 1050 and 1100. As a result, UE1 can avoid the same resource selection as UE2 by preferentially selecting a resource from
.
5. Conclusions
In this paper, we proposed two methods that contributed to the performance improvement without additional information for LTE V2X Mode 4. The proposed methods can reduce the packet collisions by using the property that reaches 0 when the resource is reselected. We evaluated the effectiveness of the proposed methods with three RRI patterns. The proposed PA method made effective use of the released resources and was highly effective in the long range. The proposed OA method avoided the overlapping selection window and was highly effective in the short range. The combination of the PA and OA methods outperformed the conventional scheme in all ranges due to the interaction of the two methods. Furthermore, no deterioration in performance was confirmed. The main contribution of this paper is to make effective use of the information contained in used in the SPS wireless resource allocation algorithm. Although various methods have been studied to improve the performance of the wireless resource allocation problem by using additional information, these methods require extensions that do not conform to the standard. In contrast, this research utilizes that have been used in standards to improve performance while maintaining compatibility. The greatest advantage of this method is that it can be introduced without modifying existing standards. We also propose two methods for utilizing , the PA and OA, and their fusion method, which always outperforms the performance of each of them alone. Our method is designed to maintain compatibility with existing standards, so it may or may not have a significant effect, but there are no disadvantages.
For future work, we are studying the effectiveness of the proposed methods in different scenarios such as effects of the channel fading and shadowing, UE density, and road model including the NLOS environment.