A Novel Random Access Algorithm for Very High Frequency Data Exchange (VDE)

: VHF Data Exchange System (VDES) is considered as an important component of the future maritime communication system by the International Maritime Organization (IMO). On the basis of existing Automatic Identification System (AIS), VDES adds the other two higher capacity subsystems: Application Specific Message (ASM) and VHF Data Exchange (VDE). The Random Access Channel (RACH) of VDE was first introduced in the International Telecommunication Union (ITU) Recommendation M.2092-0. As the slot planning of RACH in VDE is interval, which is significantly different from the continuous slot map for access algorithms in AIS, the existing slot access algorithms cannot meet the requirements of VDE. The simulation results show that the VDE slot map can reduce the normalized throughput of the existing algorithm by 39%. A novel random access algorithm called Adaptive Traffic Load Contention Resolution Diversity Slotted ALOHA (ATL-CRDSA) is proposed in this paper. The algorithm combines the load control strategy and contention resolution scheme to overcome the challenges of the new RACH of VDE. Simulation results show that ATL-CRDSA has remarkable improvement on RACH, and makes RACH very efficient and provides low latency of the packets. The insights gained from this study may be of assistance to the Media Access Control (MAC) layer design for upcoming versions of VDES standard.

safety and navigation directly. This paper mainly focuses on the random access channel access algorithm of VDE-TER. In VDE, data transmission bandwidth may be 25 kHz, 50 kHz or 100 kHz according to the channel environment 4.
From 2013, ITU and International Association of Marine Aids to Navigation and Lighthouse Authorities (IALA) have been investigating technical standards required in the development of VDES, including technical characteristics, performance requirements, and system tests. It is expected to adopt VDES resolution in 2021 or 2022. Therefore, it will lay the policy and technical foundation for the VDES application 5. So the technical specifications will be continuously amended and additional details will be supplemented according to experience requirement in implementing the technical solutions for VDES.
With the equation of VDE technical specifications, VDE functions are becoming more concrete. As stations could send data at any time in the VDE application, packet collision is a big challenge. This characteristic is similar to the conditions of S-ALOHA (Slotted-ALOHA) random access to some extent 6. Although S-ALOHA 788 has been proposed for 40 years, and its slightly improved versions are today widely used for terminal access or short packet transmissions in a shared medium, such as underwater acoustic sensor network 10, LoRaWAN networks 11, Machine-Type Communication (MTC) 11. The slot map in VDE function is interval, which is much different from the traditional S-ALOHA's continuous slot map. For the VDE RACH, we need to solve the following problems: (1) In ITU Recommendation M.2092-0, the slot map does not match the specific Logical Channel (LC). An efficient slot map based on VDES protocol is necessary. (2) The non-consecutive slot map will greatly reduce the normalized throughput of the traditional access algorithm. Therefore a new random access algorithm needs to be introduced. (3)In the VDE application, the number of mobile stations nearby the shore station is very large. Great load will increase the packet collision ratio and loss ratio, making the normalized throughput of VDE approach to 0. According to the above challenges, the existing S-ALOHA and its various improved slot access algorithms 1314 cannot meet the function and performance requirements.
Meanwhile, how to improve the utilization of slots while realizing the VDE function is also investigated in this paper. Contention Resolution Diversity Slotted ALOHA (CRDSA) is proposed in 15, which uses a replica to eliminate interference. However, in the VDE application, the normalized throughput of the CRDSA decreases rapidly when a large number of mobile stations access RACH at the same time, it cannot meet the functional requirements of VDE. Considering the VDE network with massive number of stations and the idea of load control in 1617, a novel hybrid algorithm with high normalized throughput and load control strategy is proposed in this paper, which is called Adaptive Traffic Load Contention Resolution Diversity Slotted ALOHA (ATL-CRDSA).
The main contributions of this paper include: (1) There is no specific slot map in ITU Recommendation M.2092-0 4. Through analyzing VDE functional requirements, we allocate each slot in the VDE frame and construct a new slot map. It can provide reference for future research. (2) We propose a new slot access algorithm ATL-CRDSA for VDE, which can meet the system requirements in terms of normalized throughput, packet collision ratio, packet loss ratio and system stability.
The rest of this paper is organized as follows. We discuss the related work in the second section. The frame structure, slot map and ATL-CRDSA algorithm are given in the third section. The performance bounds of ATL-CRDSA are derived and simulation results are given in the fourth section. Finally, the main conclusions are summarized, and further research is put forward.

Related Work
In the early 1970s, the first RACH media access control (MAC) protocol was proposed, which is ALOHA. It enables geographically dispersed users to share the channel to transmit data packets at random time. The vulnerable period is the slot length of a packet that is prone to collision with other packets when it is sent. The vulnerable period of ALOHA is two packet-length. Data packets will be retransmitted when they collide in accessing the RACH. ALOHA's retransmit strategy is that each transmitter waits for a random period of time and then retransmits. If a collision occurs again, the above-mentioned retransmit strategy repeats until the data packet is successfully transmitted. When the channel load is heavy, the probability of data packet collision is very high, so the channel utilization of ALOHA protocol is only 18% 18. With the continuous increase of communication capacity, how to avoid or solve collisions in access algorithms has become a crucial issue. In S-ALOHA, when the data packet arrives, it will wait for a while in the cache and send it at the beginning of the next slot. If only one packet arrives during this slot, the packet will be sent successfully in the next slot. Each packet must be transmitted within one slot. It halves the collision probability in ALOHA by dividing time into synchronized slots 19. However, when two or more data packets arrive in this slot, collision will occur. The random retransmit mechanism after S-ALOHA collision is the same as ALOHA 20. As a result, S-ALOHA shortens the vulnerable period to one packet-length, reduces the collision probability, optimizes the normalized throughput performance, and increases the maximum normalized throughput of the channel to 36.8%.
To improve the normalized throughput of RACH, transmitting multiple copies of the same packet is used to reduce the packet loss rate in Diversity Slotted ALOHA (DSA)14. Copies of data packets can be either transmitted synchronously on different frequencies or in different slots at the same frequency. It is proved that the transmission delay of multi-copy transmission is smaller and the normalized throughput of the channel is larger than that of S-AL0HA at low load conditions. However, the mechanism of DSA will greatly increase the channel load, which in VDE is very limited, so this algorithm cannot be applied in VDE directly. Duplicate data will be generated in successfully demodulated data packets, which are also a waste of channel resources. In order to assist the collision resolution, by introducing iterative interference cancellation (IIC) and frame-structure, Contention Resolution Diversity Slotted ALOHA (CRDSA) is proposed 15. In fact, the terminal will always send two copies in two randomly selected slots within the same frame. This is similar to setting the retransmission limit as 2 in 21. As the two copies have the same preamble and payload information bits, so when one copy is demodulated successfully, the recovered information will be used to cancel the interference that the other copy may generate on the other slot. Thus CRDSA greatly optimizes the performance of the system. Furthermore, Irregular Random Slotted ALOHA (IRSA) is proposed by considering the coding among the multiple replicas of a packet. The number of copies in IRSA does not exceed the random value of the declared number. Because of its large random value, the maximum normalized throughput can be up to 0.97 (packet/slot) 22. All the same, similar to the disadvantages of DSA, too many replicas will seriously increase the load of the VDE channel and reduce the utilization of channel, which is not suitable for the VDE application.
In the application for long-distance transmissions, such as satellite communication, the system is asynchronous and the frame synchronization between stations will increase the burden. Therefore, some enhanced algorithms 232425 with the asynchronous frame are proposed. As no slots are present in the frame in Contention Resolution ALOHA (CRA), the replicas of the terminals can be placed within the frame without limits, expect that replicas of a user may not completely interfere with each other. Moreover, partial interference is more probable to be decoded than complete interference 23. Based on CRA, Enhanced Contention Resolution ALOHA (ECRA) sends two copies of each packet 24. ECRA combines the advantages of CRDSA and CRA. If the interference generated by a replica is sufficiently small and the error correction code is strong enough, the packet can still be decoded correctly. The mechanism of IIC can remarkably improve the performance of ERCA. In the long transmission delay, heterogeneous transmission delay or time-varying network, an asynchronous Flipped Diversity ALOHA (AFDA) using diversity transmission and interference cancellation for MAC protocol is proposed 25, which uses Zigzag decoding to solve packet loss problem. Because of the time slots between all stations in VDE are synchronous. These asynchronous algorithms are not suitable for the RACH of VDE.
Based on the analysis of the slot map of the frame structure and performance requirements, the CRDSA is found to be most suitable for VDE. Copy packets will not increase the channel load significantly and maximize the normalized throughput. Meanwhile, we add load control based on the CRDSA to enhance the robustness of VDE and apply it in the application.

Frame Structure and Slot Map
In VDE, signals and interference are changing with time and location, so shore stations measure channel quality according to received mobile station signals and request the corresponding mobile station to use adaptive modulation and coding to maximize spectral sufficiency and normalized throughput 26. In VDE, the physical channel is divided into the upper leg and the lower leg. The upper leg is used for data transmission between shore-to-ship and ship-to-ship, and the lower leg is used for ship-to-shore data transmission. The physical channel configuration of VDE is shown in Fig. 1. VDE introduces the concept of LCs in the frame, which defines functions for a set of continuous slots 4. Each slot is allocated with one LC in one frame. The LCs are classified into data channels for data transmission and signalling channels for signalling transmission and synchronization. The connection between mobile and shore stations is session-oriented with a LC being reserved for a particular mobile station for a given time. The division of LCs is based on fixed time slots, and each channel uses a configurable number of slots at a predetermined frequency. In VDE, signalling channels are divided into Terrestrial Bulletin Board (TBB) Signalling Channels, Announcement Signalling Channel (ASC) and RACH. There are 12 LCs in VDE-TER, five of which are data channels and seven are signalling channels 26. In order to illustrate the application scenario of VDE, we present the slot map of VDE, show the available slots for RACH.

Frame Structure
According to ITU Recommendation M.2092-0, VDE frame structure is shown in Fig. 2. The length of a frame is 60 seconds, each frame is divided into 2250 slots, and each slot occupies about 26.67 milliseconds. VDE packet transmissions shall always fit into one slot. The number of bits transmitted per VDE packet shall be fixed, depending on the modulation and coding scheme used. A packet shall consist of one or multiple VDE messages, zero padding and a 4-byte Cyclic Redundancy Check (CRC). The time slots in VDES are synchronized by the Coordinated Universal Time (UTC). Six slots are combined into one Hex Slot (HS), and the cycle period of the LC is six HS 28. In VDE, 2250 slots correspond to the specific LC shown in Fig. 3. Each column represents a HS. Note that the number inside each block indicates the slot number. The yellow block in Fig. 3 is the TBB signalling channel, which occupies 18 slots 428 in a frame. The blue blocks correspond to ASC1-ASC5, which can control five parallel data transmission tasks. The red block indicates RACH, which realizes the channel access of VDE. The green block corresponds to the data channel for sending and receiving data.  Figure 3. VDE-TER ship-to-shore slot to LC mapping In the VDE application, the first three HS of each frame are allocated to the TBB signalling channel. TBB is sent by the shore station to define each Physical Channel with its VDE slot map for a control station service area. During this period, the mobile station can only receive information from the shore station, and cannot transmit any data. The HS corresponding to the same ASC number in ASC1-ASC5 handles a set of data transmission tasks respectively. The RACH resources of VDE-TER are very limited. This cyclic sorting mode in VDE can optimize the ability of data-parallel processing.
VDE-TER can realize request, resource allocations or short message transmission between shore station and mobile station in RACH, and ship-to-ship communication in the shore station service area. By monitoring the TBB, mobile stations will determine if they are within a control station service area. When a mobile station carries on the burst information transmission, the transfer starts with a Resource Request message to announce the source and destination. The following Resource Allocation message is transmitted to assign a channel to the data session. The LC is kept allocated until all fragments have been received by the mobile station and an ACK has been received or a retry limit has been exceeded. In addition, some smaller packets in communication, such as acknowledgment or paging, can also be transmitted directly on RACH.

ATL-CRDSA Algorithm
In a real environment, the number of mobile stations access to the RACH will increase greatly due to the large density of mobile stations within the range of the shore stations. Previous studies have shown that the normalized throughput of random access algorithms based on competition mechanism increases first and then decreases with the increase of load 29. For the sake of optimizing the system performance, we introduce load control in the CRDSA to keep the system normalized throughput running at a high level steadily.

CRDSA Contention Resolution
Compared to S-ALOHA and DSA, it is found that one way to improve the normalized system normalized throughput is increasing the replica of data packets. In VDE, RACH is the discontinuous HS, which means channel resources are very limited. The number of replicas will obviously increase the channel load and a higher load will greatly affect the performance. Therefore, it is not suitable to increase the normalized throughput of VDE by adding replica simply.
In CRDSA, by exploiting interference cancellation techniques, only one replica is added, the effect on the channel load is acceptable. The demodulation process confirms the signalling packet by FEC parameters and completes the decoding of the signalling packet. Valid information of two replicas in the same signalling packet contains the slot location of the corresponding replica. In fact, if one of the replicas is successfully decoded, the information about the other allows resolving the possible generated collisions. Therefore, it can recover most of the signalling packets due to collision. Fig. 4 shows an example of contention resolution in the CRDSA. In Fig. 4, the signalling packet P2 on slot 4 can be demodulated successfully, and then P3 suffered the collision of P2 on slot 3 can be recovered. Similarly, P1 and P6 can also eliminate interference. Meanwhile, we can also see the limitations of the algorithm. When two copies of the signalling packets interfere with each other, they cannot be demodulated effectively. In Fig. 4, the P4 and P5 signalling packets are in this situation. Anyway, this dual-packet iterative demodulation scheme can significantly improve normalized throughput. The performance of CRDSA is evaluated in the continuous slot system using MATLAB in Section 5. Meanwhile, the upper limit is deduced theoretically in 15, and two assumptions are made here: (1) all data packets can be demodulated correctly; (2) the collision probability of the preamble can be ignored.
When the normalized load is G and the number of iterations is Ni, normalized throughput T can be expressed as: Assuming that the transmission slot is randomly selected and will not be sent on the same time slot, the probability of the i th interference signalling packet on a given slot Sn can be obtained. As the i th packet are superimposed in slot Sn, the remaining 1 RA sn G M i  − − packets will not be present in slot Sn.
When the replica of the signalling packet is with no interference, we can obtain According to Equation (5), the first iteration can demodulate the signalling packet without interference, and the second iteration can eliminate the case of only one interference packet, and so on. It will end until the time slot with the most interference is demodulated successfully. But one case which is presented as the P4 and P5 in Fig. 4 will form deadlock. So the demodulation probability of equation (5) represents the upper limit of CRDSA.

Traffic Load Control
Load control strategy was first used in cellular networks 29, and later applied to machine to machine (M2M) communication systems. Load control can enhance robustness which is an important part of system security. In the VDE application, the number of mobile stations nearby the shore station usually is very large, and the channel load increases dramatically.
We define the maximum normalized throughput of the system Tmax. When In VDE, the shore station can monitor the channel to obtain the statistical information of RACH, estimate the ( | ) i T N G , and determine the normalized load G. According to Equation (1), we can get: And then 2 max / (1 (1 ( | )) ) In the CRDSA, the collision ratio of the signalling packet will increase and normalized throughput will decrease rapidly if the actual load is higher than Gm. In the ATL-CRDSA, we introduce an additional probability pATL, which is broadcasted by the shore station. The current signalling packet of the mobile station will be sent by probability pATL.
The probability of load control strategy is defined as: According to the ATL-CRDSA, when the load is less than Gm, the transmission probability of a mobile station does not change. Contrary, when the load exceeds Gm, the shore station broadcasts pATL, and the signalling packet will be sent by pATL, will be abandoned by (1-pATL), thus the load of the system is always near Gm.
If the probability density function is set to () t  , it can be found that the probability density function is () Applying the CRDSA to VDE can increase the normalized throughput of the RACH. Besides that, the load control strategy is introduced to ensure that the channel can maintain excellent performance in the heavy load condition, such as around the shore station.

Performance Evaluation
To illustrate the performance of ATL-CRDSA, we simulated the random access of 200 mobile stations around the shore station to the RACH, respectively obtained the performance of CRDSA in the continuous slot map environment and the VDE slot map environment, and finally simulated the performance of ATL-CRDSA in the VDE slot map environment. In order to study the performance of the algorithm, we made a detailed comparison of normalized throughput, packet collision ratio and packet loss ratio.

Normalized Throughput: T
In Fig. 6, we compare the normalized throughput of ATL-CRDSA in VDE slot map with CRDSA in continuous and VDE slot map. In the continuous slot map environment, when the optimal normalized average load of CRDSA is 1, the normalized throughput T can reach 0.93 (packet/slot). However, due to the particularity slot map of VDE, the normalized average load of the system is reduced to about 0.5, and the normalized throughput is reduced to 0.53 (packet/slot). Moreover, when the system achieves optimum performance, the increase of load will quickly destroy the behavior of the system, and the normalized throughput will decrease rapidly. The rate of decline will reach 0.67, which is much larger than that of 0.3 in the continuous slot scene. It can be observed that when the normalized average load increases to 2, the normalized throughput of CRDSA in the VDE slot map is close to 0, which completely fails to meet the requirements of the VDE application. It can be found that the normalized throughput of ATL-CRDSA can be maintained about 0.52 (packet/slot) even if the load continues to increase. From simulation results that when the load of the system reaches its optimum level, a continuous load increase will seriously damage the system performance. ATL-CRDSA is an effective solution to solve this problem.

Packet Collision Ratio
In Fig. 7, the packet collision ratio of the three circumstances over G is presented. Compared to the VDE slot map, the packet collision ratio of CRDSA in continuous slot map is significantly better but it is still worse than the ATL-CRDSA. Above the value of G = 0.6, CRDSA in continuous slot map keeps increasing and reaches about 0.9 when the G comes to 2. It is three times as much as ATL-CRDSA.

Packet Loss Ratio
In Fig. 8, for the CRDSA, in a continuous slot map system, when G = 0.8, there will be more apparent packet loss. When G = 2, the packet loss ratio will reach 0.7. In comparison for VDE slot map, when the normalized load G = 0.4, it will start to lose packets. When G = 2, the packet loss ratio will reach 1. The performance of packet loss ratio of CRDSA will be degraded due to the reduction of available slots and the interval of available slots. But the packet loss ratio will always keep at a low level for ATL-CRDSA. It can satisfy the needs of the system very well.

Conclusions
The ITU is seeking a new scheme towards digital communication applied in the maritime community. In the final version, VDE is a crucial component of VDES which can handle all the digital communication. Exploit a new algorithm for RACH in VDE can enhance system performance. In this paper, taking into consideration of the RACH's interval slot map and huge number of mobile stations, the ATL-CRDSA algorithm is proposed and in-depth analyzed for the special RACH model of VDE. It is shown that ATL-CRDSA provides reliable performance by introducing iterative interference cancellation (IIC). In practice, to overcome the impact of channel load on the system performance, the load control strategy is introduced. The simulation results show that the VDE slot map can reduce the normalized throughput of the existing algorithm by 39%. When the normalized throughput exceeds the optimal point, the normalized throughput decline rate of the VDE slot map is 2.2 times higher than that of the continuous slot map system. To fully optimize the application of ATL-CRDSA in the VDE application, the allocation of the RACH slot map for VDE needs further improvement. In this paper, we consider that data transmission in the channel will not lose byte files and there is no noise interference in the transmission environment. In practical applications, the communication environment will be more complex. For further study, we should find out the effect when introducing the channel noise in ATL-CRDSA.