Overloading Control for Machine Type Communication in LTE System

項目	內容
所屬技術領域	Now the wireless cellular communications are going to provide services for the Machine Type Communication (MTC). 　MTC represents the applications which are triggered and received by devices without the interactions of humans. Therefore, the Base Station (BS) in the future has to serve both the Human to Human (H2H) communication and the MTC. H2H communication represents the applications which are triggered and received by humans. In order to differentiate the applications, we use “User Equipment” (UE) to represent the equipment applied in H2H communication and we use “devices” to present the equipment applied in the MTC.
所欲解決之問題	It is worthy to note that devices used in MTC may have different traffic types and characters from the UEs, which are now popular in the wireless cellular communication system. Basically speaking, the differences include: 　(1) The number of devices that one BS has to serve may be several orders larger than the number of UEs that the BS serves now.  　 (2) It is possible that a burst of access attempts may come in a very short period of time, which means the MTC devices may create system overloading to the BS.
解決問題之技術手段	In order to prevent overloading condition, LTE has provided two mechanisms in the Random Access Procedure. The first is Backoff Indicator (BI) in the Random Access Response (RAR) message, which is the message broadcasted by BS. Another one is the Access Check Barring (ACB) mechanism.
對照先前技術之功效	A. Backoff Indicator (BI) 　Fig.1 plots the random access procedure. In the random access procedure, The UE selects one preamble from a set of preambles and then transmits it to the BS in the Physical random access channel (PRACH). The configuration of PRACH and preamble set is predefined and broadcasted by the BS. The BS will broadcast the Random Access Preamble IDentifier (RAPID) that it detects recently in the Random Access Response (RAR message). Therefore, UEs or devices will guess if the preamble they send is already detected by BSs by receiving the RAPIDs in the RAR message. 　 Because UEs select the preambles randomly, collisions may happen during the preamble transmission. To prevent this, a backoff indicator (BI) is attached in the RAR message, . If the UE does not receive the RAPID of the preamble that it transmits previously, it will select a backoff time . Then, the UE will step into idle state and stays as long as . BI could decrease the collision probability in the random access procedure. However, based on the LTE specification now, the maximum value of BI is only 960 ms, which means the burst of random access attempts can only postpone less than one second. The BI cannot prevent continuous large amounts of access requests. B. Access Check Barring (ACB) mechanism 　However, when a burst of UEs/devices want to access the system, the BS will trigger the ACB mechanism to prevent the overload condition. In ACB mechanism, each UE contains one fixed access class, which is distributed randomly between 0~9. When the BS is overloaded, it will bar some ACs randomly by broadcasting the ACB information. ACB information contains the following information: ACB check When the network is congested, the BS will update the ACB information to control the access attempts from the UEs. In LTE, the ACB information belongs to the system information and it is contained in the system information block Type 2 (SIB2). BS broadcasts the SIB2 every 160 ms. However, the updating of SIB2 is more dedicated. Firstly, System information normally changes only at specific radio frames whose System Frame Number is given by SFN mod N = 0, where N is configurable and defines the period between two radio frames at which a change may occur, known as the modification period. Second, the BS notifies the UEs by other SIB and by paging message before the modification period. All the UEs have to update all the system information. Therefore, all the UEs have to listen to the total latest system information, which is not only the SIB2, during the modification period. Because the notification is implemented in the paging process, the modification period is expressed as a multiple of the cell-specific default paging cycle, which may be as long as tens of seconds. The change of system information is plotted in Fig. 2. Furthermore, the devices will use the previous ACB information before it receives the latest ACB information through two modification periods. Therefore, the delay of ACB information update process will decrease the efficiency of ACB mechanism. BS still face overloading when lots of devices send access requests during the update process of ACB information. Therefore, to serve MTC devices, it is necessary to create an MTC specific overloading control mechanism in the BS.
實施方式	There are four embodiments in this document: I. Embodiment #1 　The state flow of the Embodiment#1 is plotted in Fig. 4. It includes three processes. 1. Reception of the RAR message. 2. MAU bit check. 3. MTC-ACB check 1. Reception of the RAR message 　 Firstly, the device will start a new access attempt; it will initiate a “TRAR backoff process” by triggering the TRAR. During the TRAR backoff process, the device should try to receive the RAR message. If the device does not receive the RAR message before the expiration of TRAR, the device will receive the latest MTC-ACB information and move to the MTC-ACB check procedure directly. If the device receives a RAR message before the expiration of TRAR, then the device will detect the RAR message and move to the MAU bit check process. 2. MAU bit check 　We introduce a new information bit called “MTC-ACB Update” (MAU) bit in a broadcast or multicast message. The RAR message in random access procedure is taken as an example here. The RAR message structure is plotted in Fig. 5. One bit in the sub-header, such as the reservation bit in the sub-header E/T/R/R/BI, can be the MAU bit. When the BS decides to update the MTC-ACB information, it will set the MAU bit to 1. When devices detect the MAU bit is 1, they should stay in the idle state until the next modification period. Then the device will redo the “Reception of RAR message” process. When the MTC-ACB information is updated, the BS will set the MAU bit to 0. After observing the MAU bit is set to 0, the device will start the random access procedure by using the latest MTC-ACB information.  3. MTC-ACB check 　After receiving the MTC-ACB information, the devices will perform the MTC-ACB check. In this embodiment, the MTC-ACB check reuses the LTE ACB mechanism. Furthermore, the calculation of Tmtcbarring is also the same with the Tbarring, which is shown in (1). II. Embodiment #2 　The process of embodiment#2 is plotted in Fig. 6. Fig. 6 MTC-ACB procedure (Embodiment #2) The state flow of the Embodiment#2 is also combined with three major processes. 1. Reception of the RAR message. 2. MAU bit check 3. MTC-ACB check 1. Reception of the RAR message 　 The reception of the RAR message is the same with the embodiment#1. 2. MAU bit check 　 In the embodiment#2, the ACB information is used as an example of the MTC-ACB information. Here, we use one set of MTC-ACB information to control both the ACB mechanism and MTC-ACB mechanism. We assume each device also has one Access Class and the control mechanism is defined by following rules. (a) For each AC, We brodacast the ac-barringfactor but we does not broadcast the ac-barringtime. In this condition, it is equivalent to that the ACB mechanism does not influence the UEs. In the opposite, the devices will start up its MTC-ACB mechanism based on the ac-barringfactor. We will propose a new MTC backoff time of devices in the next paragraph. (b) When we broadcast both ac-barringfactor and ac-barringtime, it means both the ACB mechanism and MTC-ACB mechanism are triggered. The UEs will follow the conventional ACB mechanism to control the UEs’ access attempts. The devices will use the same ac-barringfactor but use the MTC-backoff time in the backoff process. In this condition, the BS use one MTC-ACB information to control both the ACB mechanism and MTC-ACB mechanism. (c) When there is no broadcasting about ac-barringfactor and ac-barringtime, it means both the ACB mechanism and MTC-ACB mechanism are not triggered. 3. MTC-ACB check 　 In the embodiment#2, the MTC-ACB is the same with ACB check. Device will generate the random variable r between [0, 1]. Device will compare r with the ac-barringfactor which it receives from the latest MTC-ACB information. The device will start the random access procedure when it passes the MTC-ACB check. Otherwise, it will calculate the Tmtcbarring and start the backoff process. After the backoff process, the devices will redo the whole process by returning to the “Reception of the RAR message” process. After each fail ACB check, the device will record the number of failure MTC-ACB check (NMTCACB). The MTC-ACB procedure will be ended if the NMTCACB> Nmax_MTCACB. Otherwise, NMTCACB will be considered in the decision of Tmtcbarring. 　 In the embodiment#2, the Tmtcbarring is decided by (2). Here, the  is a random variable between [0,1].  is decided by obeying uniform distribution.  is the number of failure MTC-ACB check. The initial values of  is zero.  is the ratio to represent the buffer loading of the device,  .  is the ratio to represent the buffer loading of the BS, . To represent the loading of the cell,  can be decided by two different approaches: The information of  is obtained from the RAR message which receives in the “Reception of RAR message” process. Finally, we apply an offset, T_offset, to differentiate the MTC devices with the UEs further.  (a) If the ACB mechanism is not active => T_offset= 0 (b) If the ACB mechanism is active => T_offset= 1.3* ac-barringtime . The backoff process of Tmtcbarring after the device finishes the ACB check is plotted in Fig. 7. III. Embodiment #3 The state flow of the Embodiment#2 is also combined with three major processes: 1. Reception of the RAR message 2. MAU bit check  3. MTC-ACB check  which are plotted in Fig. 8. In the embodiment#3, the Reception of the RAR message process is the same with that of embodiment#1. The Reception of the MTC-ACB information process is the same with that of embodiment#2. However, the MTC-ACB check process is different to that in the embodiment#1 and embodiment#2. In this embodiment, the devices will also implement the MTC-ACB check. The major difference is the device will return to the Barred state when the device does not receive the RAR message successfully (the right RAPID that the device sends) in its random access procedure. The state flow chart after the fail RAR reception in the random access is also plotted in Fig. 8. Fig. 8 MTC-ACB procedure (Embodiment #3) Here, the  is a random variable between [0,1].  is decided by obeying uniform distribution.  is the number of fail MTC-ACB check.  is the number of fail RAR reception in its random access process. The initial values of  and  are zeros.  is the ratio to represent the buffer loading of the device,  .  is the ratio to represent the buffer loading of the BS,  . To represent the loading of the cell, can be decided by two different approaches : The information of  is obtained from the RAR message which receives in the random access procedure. Finally, we apply an offset, T_offset, to differentiate the MTC devices with the UEs further.  (a) If the ACB mechanism is not active => T_offset= 0 (b) If the ACB mechanism is active => T_offset= 1.3* ac-barringtime . The backoff process of Tmtcbarring after the device finishes the ACB check is plotted in Fig. 7. IV. Embodiment #4 The state flow of the Embodiment#4 is also combined with three major processes: 1. Reception of the RAR message 2. MAU bit check 3. MTC-ACB check which are plotted in Fig. 9. 　In the embodiment#4, the Reception of the RAR message process is the same with that of embodiment#1. The Reception of the MTC-ACB information process is the same with that of embodiment#2. However, the MTC-ACB check process is different to previous embodiments. Here, the devices will implement MTC-ACB check. The device will implement the backoff process until theTbarring expires. The decision of Tbarring is the same with that in (1). The devices will also start the random access procedure when it passes the MTC-ACB check. However, the devices will start the backoff period when it fails to receive the RAR message in the random access procedure. In Fig. 9, it can be observed that every time when the device fails to receive the RAR message, it will decide the value of Tmtcbarring. Then, the device will wait until the Tmtcbarring expires. When Tmtcbarringexpires, the device will start the next random access procedure if NRAR>Nmax_RAR . NRAR is the number of fail RAR reception in its random access procedure. Nmax_RAR is the maximum number of fail RAR reception that the system allows in its random access procedure. In the embodiment #4, the NRAR will affect the value of Tmtcbarring. Here, the is a random variable between [0,1]. is decided by obeying uniform distribution.  is the ratio to represent the buffer loading of the device, .  is the ratio to represent the buffer loading of the BS, . To represent the loading of the cell,  can be decided by two different approaches : The information of  is obtained from the RAR message which receives in the random access procedure. Finally, we apply an offset, T_offset, to differentiate the MTC devices with the UEs further. (a) If the ACB mechanism is not active => T_offset= 0 (b) If the ACB mechanism is active => T_offset= 1.3* ac-barringtime . The backoff process of Tmtcbarring after the device finishes the ACB check is plotted in Fig. 7.