A fast uplink access method and system for scenarios where URLLC and mMTC coexist.

Abstract

This invention discloses a fast uplink access method and system for scenarios where URLLC and mMTC coexist. The method involves: a base station predicting device status and constructing an optimization problem; jointly optimizing base station scheduling and resource allocation parameters to maximize the proportion of average successful mMTC devices while ensuring the average success rate of URLLC devices meets the target; then, the base station schedules devices based on the prediction results and optimized allocation parameters; after a device sends a packet, the device status is updated and corrected based on observation results; finally, the event status of the current time slot is estimated by back-calculating the corrected device status, and the prediction for the next time slot is then performed using a Hidden Markov Model. The prediction-based optimization scheme proposed in this invention has excellent resource allocation performance and ensures high URLLC service coverage without significantly sacrificing mMTC service coverage.

Description

Technical Field

[0001] This invention belongs to the field of uplink transmission technology for machine equipment communication services, specifically relating to a fast uplink access method and system in a scenario where URLLC and mMTC coexist. Background Technology

[0002] URLLC and mMTC are two major application scenarios for 5G. The heterogeneous coexistence of these two types of services has brought unprecedented challenges to existing system design and resource allocation schemes. Currently, the industry has proposed three categories for scenarios involving the coexistence of these two types of services: 1) URLLC and mMTC hybrid independent coexistence (URLLC-mMTC mixture); 2) scalable URLLC; and 3) critical mMTC. The main difference between these three scenarios lies in the carriers of the two types of service requirements: In the first category, the URLLC-mMTC mixture scenario, URLLC and mMTC services exist independently on two different types of devices, meaning that only the service quality requirements of the URLLC and mMTC devices need to be met separately; in the second category, the scalable URLLC scenario, it is necessary to increase the connection for some MTC devices with URLLC requirements; and in the third category, the critical mMTC scenario, URLLC and mMTC services coexist on the same device, meaning that the type of service carried by any device in this scenario varies at different times—it could be a critical URLLC service or a regular mMTC service. Therefore, in critical mMTC scenarios, dynamically, intelligently, and effectively configuring spectrum resources under different service demand levels is a challenging task.

[0003] Traditional authorized access schemes offer advantages such as a clear scheduling sequence, eliminating the need for base stations to predict device activation states. Furthermore, the base station allocates dedicated resources to each device sending a scheduling request, preventing data packet collisions. However, in critical mMTC scenarios, a large number of connected devices introduces significant signaling overhead. Additionally, URLLC services with ultra-low latency requirements may experience packet loss due to latency violations while waiting for base station authorization. Therefore, traditional authorized access schemes are unsuitable for critical mMTC scenarios. Unauthorized access schemes, where data packets are sent "on demand," can significantly reduce system overhead and latency. However, the lack of coordinated scheduling can lead to data packet collisions, which worsen with a large number of active devices, severely compromising transmission reliability. Therefore, unauthorized access schemes are also unsuitable for critical mMTC scenarios. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention proposes a fast uplink access scheme based on service type prediction for scenarios where URLLC and mMTC services coexist on the same device (critical mMTC scenario).

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a fast uplink access method in a scenario where URLLC and mMTC coexist, comprising the following steps:

[0006] S1, The base station broadcasts the reserved resource pool to all devices;

[0007] S2, the base station predicts the service type of the equipment. Based on RB resources, the service type of the equipment, the probability of successful transmission of equipment in the predicted state of URLLC service, and the probability of successful transmission of equipment in the predicted state of mMTC service, the base station maximizes the average proportion of successful mMTC equipment while ensuring that the average proportion of successful URLLC equipment meets the target. Solving the optimization problem yields a set of suboptimal solutions: the number of unlicensed transmission replications α in each time slot. t Base station resource allocation number β t and the number of base station dispatching devices and The equipment scheduling status is obtained based on the optimal solution;

[0008] S3, based on the device scheduling status, the base station sends a FUG to the active device that is predicted to be an mMTC or URLLC service, and allocates 1 or β RB resource blocks to it in the scheduling resource pool. Specifically, 1 RB resource block is allocated for mMTC service and β RB resource blocks are allocated for URLLC service.

[0009] S4. If the activated device receives 1 RB resource block, it performs data transmission on that resource block; if the activated device receives β RB resource blocks, it transmits β replica data packets; if the activated device carrying the URLLC service does not receive the FUG, it selects α RB resource blocks from the reserved resource pool for unauthorized transmission; if the activated device carrying the mMTC service does not receive the FUG, it waits for the next scheduling opportunity.

[0010] In S2, when the base station predicts the service type of the equipment, it does so based on the state of the t-1 timeslot event. The base station predicts the state of the equipment within time slot t using the HMM model. Then, resources are allocated to each device based on the prediction results, and a scheduling state is generated. The device transmits data packets using the resources allocated to it by the base station, and the base station obtains the observed status of the device within time slot t. Base station observes status With scheduling state The equipment status is corrected to obtain the corrected equipment status. Then correct the status through the equipment. The state S of the event within time slot t is obtained by reverse calculation. t * Then, this state is substituted into the HMM model to continue predicting the equipment state in the next time slot.

[0011] In S2, the base station predicts the service type of the equipment using an Hidden Markov Model (HMM) to predict the equipment's activation state. The event state is modeled as a three-state Markov process. The probability that equipment k is performing "URLLC" service in time slot t is:

[0012]

[0013] The probability that device k is performing "mMTC" service in time slot t is:

[0014]

[0015] The state of device k in time slot t is:

[0016]

[0017] random variable This indicates the state of device k within time slot t. Let P be the state sequence of N events within time slot t-1. M This represents the probability that exactly M events will activate device k to state 2. Let n be the probability that event n will activate device k to state 2 in time slot t.

[0018] In S2, when scheduling resources, the base station scheduling within time slot t... A device in mMTC state, scheduling Devices in URLLC state, among which and The following constraints must be met:

[0019]

[0020] Where, N total N represents the total number of channels that the system can provide. GF This indicates the number of channels that the system reserves for URLLC services for unlicensed transmission.

[0021] In S3 Let represent the set of device sequences whose predicted state is mMTC service. This indicates the number of devices whose predicted state is mMTC service within time slot t; Let this represent the set of device sequences whose predicted state is URLLC service. This indicates the number of devices whose predicted state is URLLC service within time slot t;

[0022] Services of the same type have the same priority. When the predicted number of devices to be activated for a certain type of service exceeds the number of devices that the system can schedule, the base station randomly selects from among them; that is, when At that time, the base station was in the collection Random selection Each device is allocated 1 RB resource; when At that time, the base station was in the collection Random selection Each device is assigned β t Each RB resource is used to obtain the resource allocation conditional probability distribution when device k is predicted to be in an mMTC service state, when device k is predicted to be in an URLLC service state, and when device k is predicted to be in an inactive state.

[0023] In S2, the known set of predicted device states The expression for the average number of devices with predicted state of URLLC within time slot t is:

[0024]

[0025] in, The reliability requirements that URLLC services should meet;

[0026] Given the set of predicted device states When the predicted state is mMTC, the average number of devices that can successfully transmit data is expressed as follows:

[0027]

[0028] in, The reliability requirements to be met for mMTC services.

[0029] In S2, the number of unlicensed transmission replications α within each time slot is jointly optimized. t Base station resource allocation number β t and the number of base station dispatching devices and Under the condition of ensuring that the average percentage of successful URLLC devices meets the target, the optimization problem is to maximize the average percentage of successful mMTC devices:

[0030]

[0031] In the formula: The average number of devices required for a successful mMTC (mMTC) η is the average number of devices for a successful URLLC. URLLC The target percentage of successful URLLC devices should be achieved;

[0032] The optimization problem is transformed into a non-convex optimization problem:

[0033]

[0034] Furthermore, all optimization variables are integers, and a set of suboptimal solutions is obtained by solving them using a discrete particle swarm optimization algorithm: the number of unlicensed transmission replications α within time slot t. t * Base station resource allocation number β t * and the number of base station dispatching devices

[0035] This invention also provides a communication system for a scenario where URLLC and mMTC coexist, including K single-antenna devices and a system configured with N t The base station has 1 antenna, and each device supports two types of services with different quality of service. The type of service carried by each device varies at different times. At any given time, device k may be in one of the following three states: a) inactive state, i.e., no data packets need to be sent; b) mMTC service state; c) URLLC service state. The system reserves resources for URLLC services. For an active device that has not received a FUG, if it is currently carrying a URLLC service, the device will perform unauthorized transmission in the reserved resources.

[0036] The fast uplink access method is as follows: The base station broadcasts a reserved resource pool to all devices;

[0037] The base station sends a FUG to the device that is predicted to be a mMTC / URLLC service, and allocates 1 (mMTC service) or β (URLLC service) RB resource blocks to it in the scheduling resource pool;

[0038] If the activating device receives 1 RB resource block, data is transmitted on that resource block; if the activating device receives β RB resource blocks, β replica data packets are transmitted.

[0039] If the activated device carrying the URLLC service does not receive the FUG, then α RB resource blocks can be selected from the reserved resource pool for unauthorized transmission.

[0040] If the activated device carrying mMTC services does not receive FUG, it will wait for the next scheduling opportunity.

[0041] When a base station predicts the service type of a device, it does so based on the state S of the t-1 timeslot event. t-1 *The base station can predict the state of the equipment within time slot t using the HMM model. Then, resources are allocated to each device based on the prediction results, and a scheduling state is generated. The device transmits data packets using the resources allocated to it by the base station, and then the base station can obtain the observed status of the device within time slot t. To avoid the accumulation of prediction errors in each time slot, the base station observes the state. With scheduling state The equipment status is corrected to obtain the corrected equipment status. Finally, the status is corrected through the equipment. The state S of the event within time slot t is obtained by reverse calculation. t * Then, this state is substituted into the HMM model to continue predicting the equipment state in the next time slot.

[0042] Compared with the prior art, the present invention has at least the following beneficial effects: The solution of the present invention reserves some resources for URLLC services. For active devices that have not received FUG, if they are currently carrying URLLC services, the devices can perform unlicensed transmission and random scheduling in the above-mentioned reserved resources. Compared with the prediction-based URLLC priority scheduling scheme, the prediction-based scheduling optimization scheme proposed in the present invention not only has higher resource utilization, but also can improve the coverage of URLLC services without seriously losing the coverage of mMTC services. Attached Figure Description

[0043] Figure 1 For fast uplink access in scenarios where URLLC and mMTC coexist.

[0044] Figure 2 This is a diagram illustrating the system resource configuration.

[0045] Figure 3 This is the revised fast uplink access solution process.

[0046] Figure 4 This is a flowchart of the prediction model.

[0047] Figure 5 This is a diagram of an HMM model.

[0048] Figure 6 Generate a model for the observed sequence.

[0049] Figure 7 This represents the average amount of resources wasted by the system under the three schemes.

[0050] Figure 8 This represents the percentage of URLLC devices that were missed by the system under the three schemes.

[0051] Figure 9 This represents the percentage of mMTC devices that were missed by the system under the three schemes.

[0052] Figure 10 The percentage of devices that successfully used URLLC under the three schemes.

[0053] Figure 11 The percentage of successful mMTC devices under the three schemes. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0056] To address the demand for massive connectivity in mMTC scenarios, a Fast Uplink Grant (FUG) access scheme is proposed. In this scheme, devices do not need to send scheduling requests to the base station; instead, the base station proactively grants permissions and allocates spectrum resources to the devices. Using FUG in critical mMTC scenarios can reduce signaling overhead and avoid data packet collisions.

[0057] Consider a critical massive machine-type communication (mMTC) scenario where URLLC and mMTC coexist. This scenario includes K single-antenna devices and a network configured with N... t A base station with one antenna. The system model is as follows: Figure 1 As shown. Each device can support two types of services with different Quality of Service (QoS), and the types of services carried by each device may differ at different times. At any given time, device k may be in one of the following three states: a) inactive state, i.e., no data packets need to be sent; b) mMTC service state; c) URLLC service state.

[0058] In traditional fast uplink access schemes, devices do not need to send scheduling requests to the base station; instead, the base station proactively sends Fast Uplink Grants (FUGs) and allocates resources. Only activated devices that receive a Fast Uplink Grant can transmit data packets; ungranted activated devices must wait for the next scheduling opportunity. Due to prediction errors, the following two types of errors can occur:

[0059] 1) The device is not activated, but the base station predicts that it is activated, resulting in a waste of resources allocated to it;

[0060] 2) The device is activated, but the base station predicts that it is inactive, causing the device to miss the scheduling opportunity and have to continue to wait for FUG, which increases the access latency.

[0061] Therefore, traditional fast uplink access schemes are not suitable for URLLC services with ultra-low latency and ultra-high reliability requirements. Thus, to ensure the stringent QoS requirements of URLLC services in scenarios where URLLC and mMTC coexist, this invention modifies the traditional fast uplink access scheme. In the modified scheme, the system reserves some resources for URLLC services. For active devices that have not received a FUG, if they are currently carrying URLLC services, the devices can perform unlicensed transmissions within the reserved resources. The system resource configuration is as follows: Figure 2 As shown.

[0062] The revised solution process is as follows: Figure 3 As shown. The specific steps are as follows:

[0063] S1: The base station broadcasts the reserved resource pool to all devices;

[0064] S2: The base station sends a FUG to the device that is predicted to be a mMTC / URLLC service, and allocates 1 (mMTC service) or β (URLLC service) RB resource blocks to it in the scheduling resource pool;

[0065] S3: If the activated device receives 1 RB resource block, data is transmitted on that resource block; if the activated device receives β RB resource blocks, β replica data packets are transmitted.

[0066] S4: If the activated device carrying the URLLC service does not receive the FUG, then select α RB resource blocks from the reserved resource pool for unauthorized transmission.

[0067] S5: If the activated device carrying mMTC services does not receive FUG, it will wait for the next scheduling opportunity.

[0068] Fast uplink authorization access solution based on business type prediction

[0069] Considering event-driven URLLC services, the activation state of each device is determined by N events. At any given time, event n may be in one of the following three states: a) the event has not occurred; b) the event has occurred but is not urgent; c) the event has occurred and is urgent. Since the state of the events and the state of the devices are unknown at the base station, while the reception state of data packets at the base station can be known through observation, the above process is modeled as a Hidden Markov Model (HMM).

[0070] Based on the state S of time slot t-1 t-1 * The base station can predict the state of the equipment within time slot t using the HMM model. Then, resources are allocated to each device based on the prediction results, and a scheduling state is generated. The device transmits data packets using the resources allocated to it by the base station, and then the base station can obtain the observed status of the device within time slot t. To avoid the accumulation of prediction errors in each time slot, the base station needs to observe the state. With scheduling state The equipment status is corrected to obtain the corrected equipment status. Finally, the status is corrected through the equipment. The state S of the event within time slot t is obtained by reverse calculation. t * Then, this state is substituted into the HMM model to continue predicting the equipment state for the next time slot. The flowchart of the prediction model is as follows: Figure 4 As shown. Detailed steps are as follows:

[0071] S1: The state S of the events within time slot t-1 is known. t-1 * Based on the state transition probability of the event Obtain the state S of the event in time slot t. t The state probability distribution, and then based on the activation probability of the device by the event. The state of device k within time slot t is predicted.

[0072] S2: The base station optimizes equipment scheduling and resource allocation based on the prediction results, and then schedules and allocates resources based on the prediction results and the optimized scheduling scheme to obtain the equipment scheduling status.

[0073] S3: Device k, within time slot t, according to... The corresponding resources are used for data packet transmission;

[0074] S4: The base station obtains the observed status of the equipment within time slot t.

[0075] S5: By and By reverse reasoning To update our understanding of the device status within time slot t;

[0076] S6: By Back-estimation of the state of events within time slot t

[0077] S7: Let t = t + 1. If t < L, proceed to step S1; otherwise, terminate.

[0078] Device activation state prediction based on hidden Markov model

[0079] The HMM model is as follows Figure 5 As shown.

[0080] The event state is modeled as a three-state Markov process. The state of event n within time slot t is represented by random variables. To indicate, among which This indicates that event n did not occur within time slot t; This indicates that event n occurs within time slot t but is not urgent; Let denot n be an event that occurs and is urgent within time slot t, and given that event n is in state i in time slot t-1, the probability of transitioning to state j in time slot t is:

[0081]

[0082] Use random variables Let represent the state of device k within time slot t, where This indicates that device k was not activated during time slot t; This indicates that device k carries mMTC services within time slot t; This indicates that device k carries URLLC services within time slot t.

[0083] The activation state of device k is determined by N events. The probability of event n activating device k in time slot t is known, denoted as:

[0084]

[0085] Device k cannot be activated if event n has not occurred.

[0086]

[0087] Device k can only be activated to either "mMTC" or "URLLC" state when event n occurs. When the event activates device k to the "mMTC" state, it is more likely; when In this case, device k is more likely to be activated in the "URLLC" state.

[0088] When t=1, the initial state probability distribution of event n in time slot t=1 is:

[0089]

[0090] The probabilities that event n will activate device k to states 0, 1, and 2 in time slot t=1 are respectively:

[0091]

[0092]

[0093]

[0094] When t≥2, event n generates its state probability distribution in time slot t based on its state transition probability from the state in time slot t-1:

[0095]

[0096] The probabilities that event n will activate device k to states 0, 1, and 2 in time slot t are as follows:

[0097]

[0098]

[0099]

[0100] Therefore, the activation probabilities of N events in time slot t for device k to be in states 0, 1, and 2 are as follows:

[0101]

[0102] When all N events activate device k to state 0, device k will be in an "inactive" state in time slot t.

[0103]

[0104] in, Let N be the state sequence of events within time slot t-1.

[0105] When at least M events activate device k to state 2, device k is in "URLLC" service during time slot t.

[0106]

[0107] Among them, P MThis represents the probability that exactly M events will activate device k to state 2.

[0108] P M The following calculation process yields the result: Selecting M events from N events yields a total of... In this case, use J M,v Let represent the set of selected event sequences in case v. Therefore, the probability of case v when exactly M events activate device k to state 2 is:

[0109]

[0110] in, Let n be the probability that event n will activate device k to state 2 in time slot t.

[0111] All The sum of the probabilities corresponding to the above cases yields the probability P. M The expression is:

[0112]

[0113] Therefore, the probability that device k is performing "URLLC" service in time slot t is:

[0114]

[0115] Therefore, the probability that device k is performing "mMTC" service in time slot t can be obtained:

[0116]

[0117] In summary, the state of device k in time slot t can be predicted as follows:

[0118]

[0119] Base station scheduling scheme

[0120] use Let represent the type of resources that device k obtains within time slot t; where, This indicates that device k did not receive RB resources in time slot t; This indicates that device k obtains one RB resource block in time slot t; This indicates that device k obtains β in time slot t. t One RB resource block.

[0121] In critical mMTC scenarios, RB resources are often limited, and it may not be possible to guarantee that all active devices can be scheduled. The base station will face the problem of selecting active devices for scheduling.

[0122] Assuming base station scheduling within time slot t A device in mMTC state, scheduling Devices in URLLC status. Among them, and The following constraints must be met:

[0123]

[0124] Where, N total N represents the total number of channels that the system can provide. GF This indicates the number of channels that the system reserves for URLLC services for unlicensed transmission.

[0125] use Let represent the set of device sequences whose predicted state is mMTC service. This represents the number of devices whose predicted state is mMTC service within time slot t; similarly, it uses... Let this represent the set of device sequences whose predicted state is URLLC service. This indicates the number of devices whose predicted state is URLLC service within time slot t.

[0126] Services of the same type have the same priority. When the predicted number of devices to be activated for a certain type of service exceeds the number of devices that the system can schedule, the base station randomly selects from among them; that is, when At that time, the base station was in the collection Random selection Each device is allocated 1 RB resource; when At that time, the base station was in the collection Random selection Each device is assigned β t RB resources.

[0127] Therefore, when device k is predicted to be in the mMTC service activation state, its resource allocation conditional probability distribution is as follows:

[0128]

[0129]

[0130]

[0131] Similarly, when device k is predicted to be in the URLLC service activation state, its resource allocation conditional probability distribution is as follows:

[0132]

[0133]

[0134]

[0135] When the predicted device k is in an inactive state, its resource allocation conditional probability distribution is as follows:

[0136]

[0137]

[0138]

[0139] Resource allocation optimization

[0140] I. Predict the success rate of device transmission for URLLC services

[0141] (1) If the resource allocation status is

[0142] The maximum transmission power of device k is P max When the device resource allocation status is The device will receive β t One RB resource, transmitting β t There are 10 identical replica packets, each with a transmit power of 100 kW. In a single-input multiple-output system, the maximum number of bits that device k can transmit on the i-th RB resource block can be approximated as:

[0143]

[0144] Where T is the length of an RB in the time domain, B0 is the bandwidth of an RB, and γ k For large-scale fading from device k to base station, g i Let Ni be the instantaneous channel gain of channel i, and N0 be the one-sided power spectral density of the noise. It is the inverse function of the Q function, e k,i Let be the transmission error probability of device k on the i-th RB.

[0145] Consider a two-state transmission model suitable for short packets. Since the maximum number of bits R that can be transmitted on an RB resource block is... k,i It concerns the instantaneous channel gain g. i Since it is an increasing function, there exists a channel gain threshold g. k th When g i ≥g k th At that time, data packets can be in 1-e k,t The probability of successful transmission is 0; otherwise, transmission fails. Let the size of the data packet be D, and let R... k,i =D, then we can solve for the device k receiving β. t Channel gain threshold g when there are RB resources β,k th for:

[0146]

[0147] For each replicated packet of device k, the successful transmission condition is: 1) The instantaneous channel gain is not less than the channel gain threshold g. k th 2) Transmission is error-free. Therefore, the probability of successful transmission of each replica packet of device k on any channel is:

[0148]

[0149] in, It is the probability density function of the instantaneous channel gain.

[0150] Due to device k transmitting β t If there are multiple identical replica packets, then the successful transmission condition is that at least one replica packet is successfully transmitted. Therefore, when the resource allocation state of device k is... At that time, its transmission success probability for:

[0151]

[0152] (2) If the resource allocation status is

[0153] If device k is not scheduled by the base station, the URLLC device can randomly select α from the reserved RB pool. t Each device performs unlicensed transmission. Similarly, it can be deduced that device k in α... t Channel gain threshold g for unlicensed transmission on RB resources α,k th for:

[0154]

[0155] Unlike base station-managed transmissions, in unlicensed transmissions, data packets sent by various devices may choose the same channel. When more than one data packet needs to be transmitted on a channel, a collision will occur, causing all data packets on that channel to fail to transmit. Therefore, for unlicensed transmission data packets to be successfully transmitted, the following three conditions must be met: 1) no data packet collisions occur; 2) the instantaneous channel gain is not less than the channel gain threshold g. α,k th 3) Error-free transmission. For device k, transmission is considered successful when at least one replica packet is successfully transmitted. Therefore, the probability of successful transmission for device k during unlicensed transmission is:

[0156]

[0157] in, The probability of a collision occurring between any replica packet during unlicensed transmission of device k.

[0158] The following is The calculation process is as follows: The target replica packet from device k will not collide with any of the other K-1 devices x in two ways: 1) Device x does not perform unlicensed transmission; 2) Device x does not select the channel where the target replica packet is located for unlicensed transmission. For any device x, the predicted probability of it performing unlicensed transmission is:

[0159]

[0160] Specifically, when device x is predicted to be in URLLC service, otherwise,

[0161] The total number of channels used for unlicensed transmission in the system is N. GF Then the probability that any other device x will not select the channel where the target replica packet is located for unlicensed transmission is:

[0162]

[0163] In summary, the probability that any replica of device k will not collide with any of the other K-1 devices is:

[0164]

[0165] The probability of a collision occurring with any replica of device k. for:

[0166]

[0167] Will Substitution From the expression, we can obtain the probability of successful transmission when device k performs unlicensed transmission:

[0168]

[0169] In summary, given the known set of predicted device states... The expression for the average number of devices with predicted state of URLLC within time slot t is:

[0170]

[0171] in, The reliability requirements that URLLC services should meet.

[0172] II. Predicting the success rate of device transmission for mMTC services

[0173] (1) If the resource allocation status is

[0174] The maximum transmission power of device k is P max When the device resource allocation status is When it receives 1 RB resource, the device will use P max The power is used for data packet transmission on this RB resource. In a single-input multiple-output system, the maximum number of bits that device k can transmit on this RB resource block can be approximated as:

[0175]

[0176] In the formula: g is the instantaneous channel gain of the channel, e k Let be the transmission error probability of device k on this RB.

[0177] Similarly, we can derive the channel gain threshold g when device k receives one RB resource. k th for:

[0178]

[0179] When g≥g k th At that time, data packets can be in 1-e k The probability of successful transmission is [not specified]; otherwise, transmission fails. Therefore, when the resource allocation state of device k is [not specified], [the transmission is successful]. At that time, the probability of successful transmission is:

[0180]

[0181] (2) If the resource allocation status is

[0182] If mMTC device k is not scheduled by the base station, the probability of successful transmission is 0, and it waits for the next base station scheduling opportunity.

[0183] In summary, given the known set of predicted device states... When the predicted state is mMTC, the average number of devices that can successfully transmit data is expressed as follows:

[0184]

[0185] in, The reliability requirements that mMTC services should meet.

[0186] 3) Constructing an optimization problem

[0187] Given a limited total spectrum resource, the number of unlicensed transmission replications α within each time slot can be jointly optimized.t Base station resource allocation number β t and the number of base station dispatching devices The goal is to maximize the average successful mMTC device ratio while ensuring that the average percentage of successful URLLC devices meets the target. The optimization problem is as follows:

[0188]

[0189] In the formula: The average number of devices required for a successful mMTC (mMTC) η is the average number of devices for a successful URLLC. URLLC The target percentage of successful URLLC devices should be achieved.

[0190] 4) Optimization problem solving

[0191] First, the optimization problem is analyzed. If the set of predictions obtained within time slot t... This is an empty set, meaning the base station predicts that no device will be activated in URLLC state within this time slot. Furthermore, the aforementioned constraint (5) is invalid. And as can be seen from the mathematical expression above, Only with optimization variables Therefore, in this case, the original optimization problem can be transformed into:

[0192]

[0193] And from From the expression, we can obtain, It is about Since it is an increasing function, the optimal solution to the above optimization problem is easily obtained as:

[0194]

[0195] The set of predictions obtained within time slot t When the set is not empty, utilize the objective function. It is about The property of the function being an increasing function allows us to rewrite the constraint (4) in the aforementioned optimization problem as follows:

[0196]

[0197] That is, when β is determined t and Once a feasible solution is found, the result can be obtained using the above formula. A feasible solution.

[0198] And because Only with the optimization variable β tThis is relevant, and when all other optimization variables are fixed, It's about β t The decreasing function, therefore the objective function Also about β t It is a decreasing function. Therefore, we can let By using β t A one-dimensional search can be performed to find the β that satisfies the conditions. t The minimum value of β, which is also the optimal solution β of the optimization problem. t * .

[0199] The aforementioned β t * Substituting into the original optimization problem, the aforementioned optimization problem can be transformed into:

[0200]

[0201] Since this problem is a non-convex optimization problem, and the optimization variables are all integers, it is solved using a discrete particle swarm optimization search algorithm. The detailed steps of this algorithm are shown in Algorithm 2.

[0202]

[0203]

[0204] By solving the above optimization problem, a suboptimal solution can be obtained: the number of unlicensed transmission replications α within time slot t. t Base station resource allocation number β t * and the number of base station dispatching devices

[0205] Error correction based on base station observation results

[0206] To prevent the accumulation of prediction errors in each time slot, the device status needs to be corrected. This update and correction process consists of two steps: the first step is to use the observation status at the base station... and the resource allocation status of the equipment. The device status is obtained by reverse engineering. The second step is to check the updated device status. Back-engineering the state of the event in the current time slot Then we can use the revised version Then continue to predict the equipment status in the next time slot.

[0207] Base station observation sequence generation

[0208] The types of data packets received by the base station from device k within time slot t are observable, using... To represent. Among them... This indicates that the base station did not receive a packet from device k within time slot t; This indicates that the base station received an mMTC data packet from device k within time slot t; This indicates that the base station received a URLLC data packet from device k within time slot t; This indicates that the base station received a packet from device k within time slot t, but it is unable to determine the type of the received data packet.

[0209] The generation model diagram of the observation sequence is shown below. Figure 6 As shown. According to the definition of a Hidden Markov Model, a device k can be given an observation sequence of length L. The generation process is described as follows:

[0210] Input: The initial state probability of event n State transition probability of event n Activation probability of event n on device k Number of base station dispatching devices Unlicensed transmission replication times α t * Base station resource allocation β t * ;

[0211] Output: Observation sequence for device k

[0212] (1) According to the distribution of the initial state of the event generate

[0213] (2) Let t = 1;

[0214] (3) According to activation probability Generate device status

[0215] (4) According to the number of base station scheduling devices In the prediction set and Select a device for scheduling and generate the resource allocation status of device k.

[0216] (5) Each device is configured according to its resource allocation status. Unlicensed transmission replication times α t * and the number of base station allocated resources β t * Data packets are transmitted to generate observation status at the base station.

[0217] (6) According to State transition probability State of generation

[0218] (7) Let t = t + 1; if t < L, proceed to step (3); otherwise, terminate.

[0219] Equipment status correction

[0220] When it is known that device k is in the observation state of the base station and resource allocation status In this case, the state of device k within time slot t can be deduced. Then its state can be predicted. Update and correct.

[0221] (1) When the observation result at the base station is If an mMTC data packet is successfully received from device k within time slot t, then it can be concluded that the device must be performing mMTC service in that time slot.

[0222] (2) When the observation result at the base station is If a URLLC data packet is successfully received from device k within time slot t, then it can be concluded that the device must be providing URLLC service in that time slot.

[0223] (3) When the observation result at the base station is The resource allocation status is or If the device receives resources but the base station does not receive packets, it indicates that the device is inactive in that time slot.

[0224] (4) When the observation result at the base station is Meanwhile, the resource allocation status is... There are three possible scenarios: a. The device is inactive and does not need to send packets, so the base station will not receive the packets; b. The device is running mMTC service, but because it is not scheduled by the base station, it cannot send packets, so the base station will not receive the packets; c. The device is running URLLC service, and because it is not scheduled by the base station, it performs unauthorized transmission, but because the unauthorized transmission fails, the base station cannot determine which device the packet came from, so the base station believes that it has not received the packet from device k.

[0225] In this case, the state of device k in time slot t can be deduced by Bayesian decision. Bayes' theorem is shown below:

[0226]

[0227] in, Let be the prior probability that device k is in the true state i in time slot t. The derivation is shown below.

[0228] When the actual status of the device is When the device is inactive, no packets need to be sent, and the observation status at the base station will always be [not specified]. Therefore:

[0229]

[0230] When the actual status of the device is At that time, the device needs to send mMTC packets, but because If the device is not scheduled by the base station, it cannot send packets; the observation result at the base station will necessarily be... Therefore:

[0231]

[0232] When the actual status of the device is At that time, the device needs to send URLLC packets, and because This indicates that the unlicensed transmission failed. The base station cannot determine which device the packet originated from, and therefore assumes that no packet was received from device k. Thus, the observation state is... Therefore:

[0233]

[0234] Therefore, when the observation result at the base station is Meanwhile, the resource allocation status is... Substituting the above calculation results into the Bayesian discriminant formula, we can obtain:

[0235]

[0236]

[0237]

[0238] In summary, based on the resource allocation status of the equipment and posterior probability The device status can be updated and corrected as follows:

[0239]

[0240] (5) When the observation result at the base station is Meanwhile, the resource allocation status is... If the error occurs, it indicates that a transmission error has occurred, preventing the base station from determining the data packet type. In this case, the state of device k within time slot t can be deduced using Bayesian decision. Bayes' theorem is shown below:

[0241]

[0242] in, Let be the prior probability that device k is in the true state i in time slot t. The derivation is shown below.

[0243] When the actual status of the device is At that time, the device is in an inactive state and does not need to send packets; the observation result at the base station is... The probability is 0. Therefore:

[0244]

[0245] When the actual status of the device is At the same time, the resource allocation status is When the device fails to transmit, the observation result at the base station is: Therefore:

[0246]

[0247] When the actual status of the device is At the same time, the resource allocation status is When the device fails to transmit, the observation result at the base station is: Therefore:

[0248]

[0249] Therefore, when the observation result at the base station is Meanwhile, the resource allocation status is... Substituting the above calculation results into the Bayesian discriminant formula, we can obtain:

[0250]

[0251]

[0252]

[0253] In summary, based on the resource allocation status of the equipment And the posterior probability can be used to correct the device state update as follows:

[0254]

[0255] (6) When the observation result at the base station is At the same time, the resource allocation status is This indicates that a transmission error in the device caused the base station to be unable to determine the data packet type. In this case, the state of device k within time slot t can be deduced using Bayesian decision. Bayes' theorem is shown below:

[0256]

[0257] in, Let be the prior probability that device k is in the true state i in time slot t.

[0258] The calculation is as follows: when the actual state of the device is At that time, the device is in an inactive state and does not need to send packets; the observation result at the base station is... The probability is 0. Therefore:

[0259]

[0260] When the actual status of the device is At the same time, the resource allocation status is When the device fails to transmit, the observation result at the base station is: Therefore:

[0261]

[0262] When the actual status of the device is At the same time, the resource allocation status is When the device fails to transmit, the observation result at the base station is: Therefore:

[0263]

[0264] Therefore, when the observation result at the base station is Meanwhile, the resource allocation status is... Substituting the above calculation results into the Bayesian discriminant formula, we can obtain:

[0265]

[0266]

[0267]

[0268] In summary, based on the resource allocation status of the equipment and posterior probability The device status can be updated and corrected as follows:

[0269]

[0270] The above updates and corrections to the device status are summarized in the table below:

[0271] Table 1 Equipment Status Correction

[0272]

[0273]

[0274] Event State Estimation

[0275] When the corrected device state A is known 1:t At that time, the state of events within the current time slot can be estimated by using Bayesian decision, that is:

[0276]

[0277] The event state can then be determined as follows:

[0278]

[0279] Furthermore, since directly calculating the aforementioned posterior probabilities is highly complex, and the denominators are all the same, the maximum a posteriori probability decision can be equivalent to the maximum joint probability decision, that is:

[0280]

[0281] The joint probability P(A) is calculated using the forward algorithm. 1:t ,S t This can significantly reduce computational complexity. The forward algorithm defines this joint probability as the "forward probability," that is, the state sequence of the device in the first t time slots is A. 1:t And the state sequence of the event in time slot t is S t The probability of forward probability. The essence of the forward algorithm is to use the idea of ​​dynamic programming, calculate the local forward probability, and then use the path structure of the state sequence to recursively apply the forward probability to the global probability.

[0282] The recursive formula for the forward probability is:

[0283]

[0284] When the time slot event state S is obtained... t * When the time slot t+1 is reached, the device status can be predicted by the HMM model, and then the next round of base station scheduling and resource allocation can be carried out.

[0285] The system simulation parameters are shown in the table below.

[0286] Table 2 System Simulation Parameter Settings

[0287]

[0288]

[0289] If a base station allocates spectrum resources to a device, but the device is actually inactive, then the resources allocated to the inactive device are considered wasted. Figure 7 The figure shows the curves of the average number of wasted resource blocks over time under three different schemes. The optimized scheme is the one proposed in this invention, which jointly optimizes the scheduling of devices and resource allocation based on prediction results. Comparison Scheme 1 is the Random Scheduling (RS) scheme, where the base station does not predict the activation state of devices and directly grants and allocates resources randomly to all devices. Comparison Scheme 2 is the prediction-based URLLC priority scheduling scheme, where the base station obtains two sets of device sequences with predicted states of URLLC and mMTC, respectively. To ensure the strict latency requirements of URLLC, in this scheme, the base station first schedules the high-priority URLLC with limited resources. If there are still resources remaining after all URLLCs are scheduled, then devices are randomly selected from the set of devices with the predicted state of mMTC for scheduling. As shown in the figure, the scheme proposed in this invention has the fewest wasted resources, the prediction-based URLLC priority scheduling scheme has the second best system resource utilization performance, and the random scheduling scheme has the most wasted resources. Comparing the optimized scheme of this invention with the random scheduling scheme, it can be seen that the base station can significantly reduce the waste of spectrum resources by granting uplink authorization to the equipment based on the prediction results; comparing the optimized scheme of this invention with the URLLC priority scheduling scheme, it can be seen that optimizing the base station scheduling scheme based on the prediction results of the equipment service type can further improve the resource utilization of the system.

[0290] as follows Figure 8The figure shows the ratio of the number of URLLC devices missed to the total number of URLLC devices under the three schemes mentioned above, changing over time. As can be seen from the figure, the proportion of missed URLLC devices is lowest under the prediction-based URLLC priority scheme. The proportion of missed URLLC devices in the proposed optimization scheme is slightly higher than that in the URLLC priority scheduling scheme, while the proportion of missed URLLC devices is highest in the random scheduling scheme. The higher proportion of missed URLLC devices in the proposed optimization scheme is due to two main reasons. Firstly, in the proposed optimization scheme, the system reserves some dedicated resources for URLLC for unlicensed transmission, resulting in fewer available scheduling resources compared to the URLLC priority scheduling scheme, which may increase the probability of missed URLLC transmission. Secondly, in the URLLC priority scheduling scheme, scheduling resources are preferentially allocated to URLLC devices. However, in the proposed optimization scheme, since URLLC has dedicated resources, the base station may allocate the saved scheduling resources to mMTC by appropriately missing some URLLC packets, and the missed URLLC devices can be compensated through unlicensed transmission. Comparing the reserved URLLC priority scheme and the random scheduling scheme, the number of available scheduling resources is equal for both. However, the proportion of URLLC devices that are missed by the reserved URLLC priority scheme is significantly lower than that of the random scheduling scheme. This result illustrates the effectiveness and necessity of base station prediction of device service types.

[0291] as follows Figure 9 The figure shows the curves of the percentage of missed mMTC devices under the three different schemes mentioned above. As can be seen from the figure, the random scheduling scheme has the highest percentage of missed mMTC devices, followed by the proposed optimized scheme. The prediction-based URLLC priority scheduling scheme performs best in this metric. The proposed optimized scheme has a slightly higher percentage of missed mMTC devices than the URLLC priority scheduling scheme. This is because the system in the proposed optimized scheme reserves dedicated resources for URLLC, and the number of resources available for scheduling by the base station is less than that in the URLLC priority scheduling scheme, resulting in a higher percentage of missed mMTC devices under the proposed optimized scheme. Comparing the proposed optimized scheme with the random scheduling scheme, although the available scheduling resources in the proposed optimized scheme are less than those in the random scheduling scheme, the base station can achieve a lower percentage of missed mMTC devices by scheduling according to the prediction results of the device service type.

[0292] as follows Figure 10The figure shows the curves of the percentage of successful URLLC devices over time under the three different schemes mentioned above. As can be seen from the figure, the scheme proposed in this invention has the highest percentage of successful URLLC devices, reaching approximately 97%; the prediction-based URLLC priority scheduling scheme is second, with a percentage of successful URLLC devices of approximately 90%; the random scheduling scheme has the worst performance, with a percentage of successful URLLC devices of only about 50%. Comparing the two prediction-based scheduling schemes (optimized scheduling scheme and URLLC priority scheduling scheme) with the random scheduling scheme, it can be seen that scheduling devices based on the predicted service type of the device can significantly improve the percentage of successful URLLC devices, with an improvement of up to about 80%. Comparing the proposed optimized scheduling scheme with the URLLC priority scheduling scheme, although the percentage of missed URLLC devices in the optimized scheme is slightly higher than that in the URLLC priority scheduling scheme, its percentage of successful URLLC devices is still about 7% higher. This is because in the optimized scheme, missed URLLC can be transmitted without authorization using its dedicated resources, while the URLLC priority scheduling scheme does not set up a compensation mechanism for missed URLLC devices. The simulation results demonstrate that reserving dedicated resources for URLLC as a compensation mechanism and jointly optimizing the scheduling scheme and resource allocation based on prediction results can further improve the success rate of URLLC. In summary, the performance advantages of the optimized scheduling scheme proposed in this invention mainly come from the following three aspects: 1) The base station schedules based on the prediction results of the device's service type; 2) Reserving some dedicated resources for URLLC for unlicensed transmission serves as a compensation mechanism for missed URLLC transmissions; 3) Jointly optimizing the base station scheduling and resource allocation parameters.

[0293] as follows Figure 11 The figure shows the curves of the percentage of successful mMTC devices over time under the three different schemes mentioned above. As can be seen from the figure, the prediction-based URLLC priority scheduling scheme has the highest percentage of successful mMTC devices, around 65%; the proposed optimized scheduling scheme is second, with a percentage of successful mMTC devices of approximately 52%; and the random scheduling scheme has the lowest percentage of successful mMTC devices, approximately 50%. The proposed optimized scheme performs worse than the URLLC priority scheduling scheme in terms of the percentage of successful mMTC devices. This is because in the proposed scheme, the system allocates dedicated resources for URLLC, thus reducing the number of scheduling resources available to the base station, resulting in a loss of successful mMTC devices. Considering the performance of the percentage of successful URLLC devices, it can be concluded that the performance advantage of the proposed optimized scheme in terms of the percentage of successful URLLC devices comes at the cost of a lower percentage of successful mMTC devices.

[0294] In summary, compared with random scheduling schemes and prediction-based URLLC priority scheduling schemes, the prediction-based optimized scheduling scheme proposed in this invention not only has higher spectrum resource utilization, but also can significantly improve the coverage of URLLC services without severely sacrificing mMTC service coverage.

Claims

1. A fast uplink access method for scenarios where URLLC and mMTC coexist, characterized in that, Includes the following steps: S1, The base station broadcasts the reserved resource pool to all devices; S2, the base station predicts the service type of the equipment. Based on RB resources, the service type of the equipment, the probability of successful transmission of equipment in the predicted state of URLLC service, and the probability of successful transmission of equipment in the predicted state of mMTC service, an optimization problem is constructed: Maximize the average percentage of successful mMTC devices while ensuring that the average percentage of successful URLLC devices meets the target. Solving this optimization problem yields a set of suboptimal solutions: the number of unlicensed transmission replications α in each time slot. t Base station resource allocation number β t and the number of base station dispatching devices and The equipment scheduling status is obtained based on the suboptimal solution; S3, based on the device scheduling status, the base station sends a FUG to the active device that is predicted to be an mMTC or URLLC service, and allocates 1 or β RB resource blocks to it in the scheduling resource pool. Specifically, 1 RB resource block is allocated for mMTC service and β RB resource blocks are allocated for URLLC service. S4. If the activated device receives 1 RB resource block, it performs data transmission on that resource block; if the activated device receives β RB resource blocks, it transmits β replica data packets; if the activated device carrying the URLLC service does not receive the FUG, it selects α RB resource blocks from the reserved resource pool for unauthorized transmission; if the activated device carrying the mMTC service does not receive the FUG, it waits for the next scheduling opportunity.

2. The fast uplink access method in a scenario where URLLC and mMTC coexist as described in claim 1, characterized in that, In S2, when the base station predicts the service type of the equipment, it does so based on the state of the t-1 timeslot event. The base station predicts the state of the equipment within time slot t using the HMM model. Then, resources are allocated to each device based on the prediction results, and a scheduling state is generated. The device transmits data packets using the resources allocated to it by the base station, and the base station obtains the observed status of the device within time slot t. Base station observes status With scheduling state The equipment status is corrected to obtain the corrected equipment status. Then correct the status through the equipment. The state S of the event within time slot t is obtained by reverse calculation. t * Then, this state is substituted into the HMM model to continue predicting the equipment state in the next time slot.

3. The fast uplink access method in a scenario where URLLC and mMTC coexist as described in claim 2, characterized in that, In S2, the base station predicts the service type of the equipment using an Hidden Markov Model (HMM) to predict the equipment's activation state. The event state is modeled as a three-state Markov process. The probability that equipment k is performing "URLLC" service in time slot t is: The probability that device k is performing "mMTC" service in time slot t is: The state of device k in time slot t is: random variable This indicates the state of device k within time slot t. Let P be the state sequence of N events within time slot t-1. M P represents the probability that exactly M events will activate device k to state 2. t (nk) (2) is the probability that event n will activate device k to state 2 in time slot t.

4. The fast uplink access method in a scenario where URLLC and mMTC coexist as described in claim 2, characterized in that, In S2, when scheduling resources, the base station scheduling within time slot t... A device in mMTC state, scheduling Devices in URLLC state, among which and The following constraints must be met: Where, N total N represents the total number of channels that the system can provide. GF This indicates the number of channels that the system reserves for URLLC services for unlicensed transmission.

5. The fast uplink access method in a scenario where URLLC and mMTC coexist as described in claim 2, characterized in that, In S3 Let represent the set of device sequences whose predicted state is mMTC service. This indicates the number of devices whose predicted state is mMTC service within time slot t; Let this represent the set of device sequences whose predicted state is URLLC service. This indicates the number of devices whose predicted state is URLLC service within time slot t; Services of the same type have the same priority. When the predicted number of devices to be activated for a certain type of service exceeds the number of devices that the system can schedule, the base station randomly selects from among them; that is, when At that time, the base station was in the collection Random selection Each device is allocated 1 RB resource; when At that time, the base station was in the collection Random selection Each device is assigned β t Each RB resource is used to obtain the resource allocation conditional probability distribution when device k is predicted to be in an mMTC service state, when device k is predicted to be in an URLLC service state, and when device k is predicted to be in an inactive state.

6. The fast uplink access method in a scenario where URLLC and mMTC coexist as described in claim 2, characterized in that, In S2, the known set of predicted device states The expression for the average number of devices with predicted state of URLLC within time slot t is: in, The reliability requirements that URLLC services should meet; Given the set of predicted device states When the predicted state is mMTC, the average number of devices that can successfully transmit data is expressed as follows: in, The reliability requirements to be met for mMTC services.

7. The fast uplink access method in a scenario where URLLC and mMTC coexist as described in claim 2, characterized in that, In S2, the number of unlicensed transmission replications α within each time slot is jointly optimized. t Base station resource allocation number β t and the number of base station dispatching devices and Under the condition of ensuring that the average percentage of successful URLLC devices meets the target, the optimization problem is to maximize the average percentage of successful mMTC devices: 1≤α t ≤N GF 1≤β t ≤N total -N GF In the formula: The average number of devices required for successful mMTC. The average number of devices for a successful URLLC. η URLLC The target percentage of successful URLLC devices should be achieved; The optimization problem is transformed into a non-convex optimization problem: 1≤α t ≤N GF Furthermore, all optimization variables are integers, and a set of suboptimal solutions is obtained by solving them using a discrete particle swarm optimization algorithm: the number of unlicensed transmission replications α within time slot t. t * Base station resource allocation number β t * and the number of base station dispatching devices 8. A communication system for a scenario where URLLC and mMTC coexist, characterized in that, To implement the fast uplink access method for the coexistence scenario of URLLC and mMTC as described in any one of claims 1-7, it includes K single-antenna devices and a device configured with N t The base station has 1 antenna, and each device supports two types of services with different quality of service. The type of service carried by each device varies at different times. At any given time, device k may be in one of the following three states: a) inactive state, i.e., no data packets need to be sent; b) mMTC service state; c) URLLC service state. The system reserves resources for URLLC services. For an active device that has not received a FUG, if it is currently carrying a URLLC service, the device will perform unauthorized transmission in the reserved resources.

9. The communication system for the coexistence of URLLC and mMTC as described in claim 8, characterized in that, The fast uplink access method is as follows: The base station broadcasts a reserved resource pool to all devices; The base station sends a FUG to the device that is predicted to be a mMTC / URLLC service, and allocates 1 (mMTC service) or β (URLLC service) RB resource blocks to it in the scheduling resource pool; If the activating device receives 1 RB resource block, data is transmitted on that resource block; if the activating device receives β RB resource blocks, β replica data packets are transmitted. If the activated device carrying the URLLC service does not receive the FUG, then α RB resource blocks can be selected from the reserved resource pool for unauthorized transmission. If the activated device carrying mMTC services does not receive FUG, it will wait for the next scheduling opportunity.

10. The communication system for the coexistence of URLLC and mMTC as described in claim 8, characterized in that, When a base station predicts the service type of a device, it does so based on the state S of the t-1 timeslot event. t-1 * The base station can predict the state of the equipment within time slot t using the HMM model. Then, resources are allocated to each device based on the prediction results, and a scheduling state is generated. The device transmits data packets using the resources allocated to it by the base station, and then the base station can obtain the observed status of the device within time slot t. Base station observes status With scheduling state The equipment status is corrected to obtain the corrected equipment status. Finally, the status is corrected through the equipment. The state S of the event within time slot t is obtained by reverse calculation. t * Then, this state is substituted into the HMM model to continue predicting the equipment state in the next time slot.

Images