Adaptive modulation and coding method for deterministic low latency retransmission

Abstract

The application discloses a kind of adaptive modulation coding methods for deterministic low-latency retransmission.In order to optimize average latency under the premise of meeting deterministic latency constraints, the modulation coding scheme of initial transmission queue and retransmission queue is dynamically selected. In order to solve the problem of inter-slot coupling of the queue, a multi-step Q-learning method based on updating system is proposed to reacquire the value function of the system when updating event is triggered. A virtual queue corresponding to the deterministic latency constraint is constructed, and a cost function based on the virtual queue and the actual queue is designed for value function learning. Based on the obtained value function, the system determines the modulation coding scheme to be used according to the current queue length. The adaptive modulation coding scheme proposed can meet the deterministic latency constraint and minimize the average latency. The application can be used for modulation coding scheme design considering retransmission in transmission network to reduce the average latency under the premise of meeting the deterministic latency constraint.

Description

Technical Field

[0001] The present invention relates to the field of wireless communications, and in particular to an adaptive modulation and coding method for deterministic low-delay retransmission. Background Art

[0002] Ultra-Reliable and Low Latency Communications (URLLC) is an emerging application scenario. It aims to provide reliable technical guarantees for latency-sensitive services such as industrial automation, Internet of Vehicles, and Tactile Internet. Specific performance requirements vary in different scenarios. Taking automation control in industrial environments as an example, end-to-end latency is required to be less than 10ms and reliability is required to be greater than 99.999%. In URLLC scenarios, data packets are generally considered to be short packets (32-200 bytes). In order to meet such stringent latency requirements, data must be transmitted to the recipient quickly and error-free with high levels of integrity, authenticity, and confidentiality, making URLLC design a challenging task.

[0003] Adaptive Modulation and Coding (AMC) is a widely adopted physical layer mechanism. Due to the random nature of wireless channels, the long-term use of a single modulation and coding scheme often results in communication resource utilization failing to keep pace with dynamic channel changes, leading to resource shortages or waste. AMC dynamically adjusts the modulation and coding scheme (MCS) by evaluating the current channel state, thereby improving resource utilization efficiency.

[0004] Hybrid Automatic Repeat Request (HARQ) is a widely used retransmission mechanism. Data transmission errors are inevitable in wireless communications due to factors such as noise, interference, and channel fading. By combining forward error correction (FEC) and automatic repeat requests (ARQ), HARQ provides error detection and correction capabilities, and timely retransmissions when error correction mechanisms fail, improving system reliability.

[0005] The design of adaptive modulation and coding for deterministic low-latency retransmission faces the following two challenges. First, traditional AMC and HARQ schemes cannot meet today's deterministic low-latency requirements. Throughput-optimized AMC is often accompanied by a large number of retransmissions, and the round-trip delay (RTT) caused by retransmissions inevitably accounts for a part of the total delay, which puts great pressure on meeting deterministic delay. The second is the coupling relationship caused by cascaded queues. Previous work usually gave retransmitted data the highest priority / lowest MCS to minimize multiple retransmissions, but this did not take into account the comprehensive factors of the initial transmission and retransmission queues, which would drag down the overall system delay performance, which is especially important in URLLC services. In a cascaded queue system, the amount of errors in the initial transmission data arrives as the retransmission queue, forming a complex queue coupling relationship, which poses a challenge to optimizing the total delay. Summary of the Invention

[0006] To overcome the shortcomings of the existing technology, the present invention aims to provide an adaptive modulation and coding method for deterministic low-latency retransmission. To optimize average latency while satisfying deterministic latency constraints, the modulation and coding schemes for the initial and retransmission queues are dynamically selected. To address the incompatibility between the time slot coupling of the queues and the traditional Lyapunov method, an update system framework is proposed, based on which reinforcement learning is used to optimize the drift penalty. This proposed adaptive modulation and coding scheme satisfies deterministic latency constraints and minimizes average latency.

[0007] An adaptive modulation and coding method for deterministic low-latency retransmission, comprising the following steps:

[0008] 1. Obtain the following information: data arrival rate λ, time slot length τ, round-trip delay D RTT ; Number of optional modulation and coding schemes M, deterministic delay constraint of the service:

[0009]

[0010] Where D(t) is the delay at time t, d is the upper limit of delay, and ε is the tolerable probability of violating the upper limit of delay;

[0011] 2. Initialize the queue status on the user device side: initial queue length Q f =0, retransmission queue length Q r =0, initialize update event count r=-1;

[0012] 3. Each time slot determines whether the current queue meets Q f =0 and Q r = 0, if yes, then trigger the update event, record r = r + 1. Re-obtain the value function of the system where χQ =(Q f ,Q r ) is the queue status, is the size of the single queue state space, A=(m f ,m r ) are the modulation and coding schemes of the initial data packet and the retransmitted data packet respectively. If r = 0, the optimal estimation value table is randomly initialized; if r> 0, the value estimation J learned in the previous update frame is adopted (r-1) As the initial value to speed up the convergence;

[0013] Based on the initial value, the value function J is updated through multi-step Q learning. In each iteration, the current strategy is first obtained Then for all state-action pairs Perform multi-step continuous sampling:

[0014]

[0015] And update synchronously through the following formula:

[0016]

[0017] in is a decreasing step sequence, g(χ Q ,A)={g(t)|χ Q ,A}, g(t) is in χ Q The instantaneous cost of adopting the modulation and coding scheme selection strategy A in the queue state until the convergence condition ‖J is met i+1 -J i ‖ ∞,sp ≤ξ, the value function after convergence is recorded as J (r) ;

[0018] 4) In each time slot, the value function J obtained in step 3) is calculated. (r) , the modulation and coding scheme used for the initial data packet and the retransmitted data packet is determined based on the current queue length

[0019]

[0020] The cost function g(t) is designed as follows:

[0021] 3.1) To meet the deterministic delay constraint, a virtual queue Q is constructed v (t), which satisfies the following evolution process:

[0022]

[0023] Where O(t) is the delay violation, and O is the initial queue delay violation.f (t), the retransmission queue delay violation amount O r (t) and the retransmission failure data amount b r (t) the sum of the three;

[0024] O(t) = O f (t) + O r (t) + b r (t)

[0025] the initial queue delay violation amount O f (t) can be calculated as follows:

[0026]

[0027] where x(t) is the queue scheduling factor, taking the retransmission priority strategy, i.e. x(t) = 1{Q r (t) > 0} e {0, 1}, r(m f (t)) is the data transmission amount when the m f (t) modulation coding scheme is taken, and a(t) is the data arrival amount; denotes the initial queue threshold value corresponding to the delay violation;

[0028] the retransmission queue delay violation amount O r (t) can be calculated as follows:

[0029]

[0030] where r(m r (t)) is the data transmission amount when the m r (t) modulation coding scheme is taken, b f (t) is the error data amount generated by the initial transmission, denotes the retransmission queue threshold value corresponding to the delay violation.

[0031] 3.2) the cost function g(t) is

[0032]

[0033] where t (r) is the start point of the rth update frame, W is the weighting factor of the average delay penalty, which is set to an appropriate size so that the average delay is fully valued and the deterministic delay constraint is met, and D(t) is the delay at time t

[0034]

[0035] where λ is the data arrival rate, α is the retransmission proportion, is the retransmission rate, and α and can be obtained by statistics.

[0036] Beneficial effects of the present invention:

[0037] The present invention designs an adaptive modulation and coding scheme for deterministic low-latency retransmission, follows the idea of ​​comprehensive consideration of average delay and violation probability, dynamically selects the modulation and coding scheme for the initial transmission queue and the retransmission queue, and cleverly utilizes the independent and identically distributed characteristics of the update system to combine the Markov decision process and Lyapunov optimization theory. It can reduce the average delay while satisfying the deterministic delay constraint, and obtain near-optimal performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a performance comparison chart of the average delay between the algorithm proposed in the present invention and three comparison algorithms under fixed channel conditions and different average data arrival rates.

[0039] Figure 2 This is a performance comparison chart of the algorithm proposed in the present invention and three comparative algorithms in terms of deterministic delay violation rate under fixed channel conditions and different average data arrival rates.

[0040] Figure 3 : is a performance comparison chart of the average delay between the algorithm proposed in the present invention and three comparison algorithms under different channel conditions with a fixed average data arrival rate.

[0041] Figure 4 : is a performance comparison chart of the algorithm proposed in the present invention and three comparison algorithms in terms of deterministic delay violation rate under different channel conditions with a fixed average data arrival rate. DETAILED DESCRIPTION

[0042] The present invention is further described below with reference to the accompanying drawings and examples. It is worth noting that the following examples are only used to illustrate the present invention and are not used to limit the present invention.

[0043] An adaptive modulation and coding method for deterministic low-latency retransmission, the steps are as follows:

[0044] Step 1: Obtain the following information: data arrival rate λ, time slot length τ, round-trip delay D RTT ; Number of optional modulation and coding schemes M, deterministic delay constraint of the service:

[0045]

[0046] Where D(t) is the delay at time t, d is the upper limit of delay, and ε is the tolerable probability of violating the upper limit of delay;

[0047] Step 2: Initialize the queue state on the user device: initial queue length Q f =0, retransmission queue length Qr =0, initialize update event count r=-1;

[0048] Step 3: Mathematically model the queue delay based on the update system and construct an optimization problem:

[0049]

[0050]

[0051] in is the accumulated delay cost and delay violation cost in the rth update frame, T (r) =t (r+1) -t (r) is the duration of the update frame; Ω (r) is the modulation and coding scheme strategy in the rth update frame.

[0052] D(t) is the delay at time t

[0053]

[0054] Where λ is the data arrival rate, α is the retransmission ratio, is the retransmission rate, α and It can be obtained through statistics.

[0055] O(t) is the delay violation, and O is the initial queue delay violation f (t), retransmission queue delay violation O r (t) and the amount of data that failed to be retransmitted b r (t)The sum of three items.

[0056] O(t)=O f (t)+O r (t)+b r (t)

[0057] Initial queue delay violation O f (t) can be calculated as follows:

[0058]

[0059] Where x(t) is the queue scheduling factor, and the retransmission priority strategy is adopted, that is, x(t)=1{Q r (t)>0}∈{0,1},r(m f (t)) is to take the mth f (t) The amount of data transmitted when the modulation and coding scheme is used, a(t) is the amount of data arriving; Indicates the initial queue threshold corresponding to the delay violation.

[0060] Retransmission queue delay violation Or (t) can be calculated as follows:

[0061]

[0062] Where r(m r (t)) is to take the mth r (t) The amount of data transmission when using the modulation and coding scheme, b f (t) is the amount of erroneous data generated in the initial transmission, Indicates the retransmission queue threshold corresponding to the delay violation.

[0063] In order to meet the deterministic delay constraint, a virtual queue Q is constructed. v (t), which satisfies the following evolution process:

[0064]

[0065] We obtain the solution to the problem by minimizing the virtual queue drift plus delay penalty on the update frame, and use this as the cost function for subsequent multi-step Q learning:

[0066]

[0067] Step 4: In each time slot, determine whether the current queue meets Q f =0 and Q r = 0, if yes, then trigger the update event, record r = r + 1. Re-obtain the value function of the system where χ Q =(Q f ,Q r ) is the queue status, is the size of the single queue state space, A=(m f ,m r ) are the modulation and coding schemes of the initial data packet and the retransmitted data packet respectively. If r = 0, the optimal estimation value table is randomly initialized; if r> 0, the value estimation J learned in the previous update frame is adopted (r -1) As the initial value to speed up the convergence.

[0068] Based on the initial value, the value function J is updated through multi-step Q learning. In each iteration, the current strategy is first obtained Then for all state-action pairs Perform multi-step continuous sampling:

[0069]

[0070] And update synchronously through the following formula:

[0071]

[0072] in is a decreasing step sequence, g(χ Q ,A)={g(t)|χ Q ,A}, g(t) is in χ Q The instantaneous cost of adopting the modulation and coding scheme selection strategy A in the queue state. Until the convergence condition ‖J is met i+1 -J i ‖ ∞,sp ≤ξ, the value function after convergence is recorded as J (r) .

[0073] Step 5: In each time slot, calculate the value function J obtained in step 3. (r) , the modulation and coding scheme used for the initial data packet and the retransmitted data packet is determined based on the current queue length

[0074]

[0075] As a better implementation example, assuming that the parameters are not used as adjustment variables, there are M = 5 different modulation and coding schemes, the time slot length τ = 0.25ms, and the round-trip delay D RTT =1ms. The upper limit of delay d = 2ms, and the tolerable probability of violating the upper limit of delay ε = 10 -4 The computer simulation was run 1000 times and the average value was taken as the final result. Three comparison methods were introduced: the first is the slow AMC algorithm, which meets the average target error rate for large-scale fading; the second is the fast AMC algorithm, which considers small-scale fading to ensure that the target error rate can be met at the worst; and the third is the throughput optimization algorithm, which ranks the effective throughput of all modulation and coding schemes from high to low and selects the one with the highest throughput.

[0076] Figure 1 、 Figure 2 The following are performance comparison diagrams of the algorithm proposed in this invention and three comparison algorithms in terms of average delay and deterministic delay violation rate under fixed channel conditions and different average data arrival rates.

[0077] Figure 3 、 Figure 4 The performance comparison diagrams of the algorithm proposed in the present invention and three comparison algorithms in terms of average delay and deterministic delay violation rate under fixed average data arrival rate and different channel conditions are shown.

[0078] For different average data arrival rates, such as Figure 1 、 Figure 2As shown, the algorithm proposed in this paper shows a significant advantage in average latency over the three comparison algorithms as the data arrival rate increases. This means that the performance advantage of the algorithm proposed in this paper is even more pronounced under higher loads. Furthermore, the algorithm proposed in this paper can effectively meet the deterministic low-latency constraint under various load conditions.

[0079] For different channel conditions, such as Figure 3 、 Figure 4 As shown, the algorithm proposed in this paper shows a significant advantage in average latency over the three comparison algorithms as the average received signal-to-noise ratio decreases. This means that the performance advantage of the algorithm proposed in this paper is even more pronounced in poor channel conditions. Furthermore, the algorithm proposed in this paper can effectively meet the deterministic low-latency constraint under various channel conditions.

[0080] Based on the above performance comparison, the present invention dynamically selects the modulation and coding scheme for the initial and retransmission queues to achieve a deterministic, low-latency modulation and coding scheme that takes retransmissions into account. It constructs a virtual queue that satisfies deterministic latency constraints and designs a cost function based on the virtual and actual queues for value function learning. A multi-step Q-learning method based on the update system is proposed to retrieve the system's value function when an update event is triggered. The modulation and coding scheme is then determined based on the current queue length, achieving the goal of reducing average latency while satisfying deterministic latency constraints.

[0081] The above-described embodiments merely illustrate several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that a person skilled in the art would be able to make numerous modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. An adaptive modulation and coding method for deterministic low-latency retransmission, characterized in that: The steps include: 1) Obtain the following information: data arrival rate λ, time slot length τ, round-trip delay D RTT ; Number of optional modulation and coding schemes M, deterministic delay constraint of the service: Where D(t) is the delay at time t, d is the upper limit of delay, and ε is the tolerable probability of violating the upper limit of delay; 2) Initialize the queue state at the user device: initial queue length Q f =0, retransmission queue length Q r =0, initialize update event count r=-1; 3) Each time slot determines whether the current queue meets Q f =0 and Q r = 0, if yes, then trigger the update event, record r = r + 1; re-obtain the value function of the system where χ Q =(Q f ,Q r ) is the queue status, is the size of the single queue state space, A=(m f ,m r ), where m f is the modulation and coding scheme of the initial data packet, m r is the modulation and coding scheme of the retransmitted data packet; if r = 0, the optimal estimation value table is randomly initialized; if r> 0, the value estimation J learned in the previous update frame is adopted (r-1) As the initial value to speed up the convergence; Based on the initial value, the value function J is updated through multi-step Q learning; In each iteration, the current strategy is first derived Then for all state-action pairs Perform multi-step continuous sampling: And update synchronously through the following formula: in is a decreasing step sequence, g(χ Q ,A)={g(t)|χ Q ,A}, g(t) is in χ Q The instantaneous cost of adopting the modulation and coding scheme selection strategy A in the queue state until the convergence condition ‖J is met i+1 -J i ‖ ∞,sp ≤ξ, the value function after convergence is recorded as J (r) ; 4) In each time slot, the value function J obtained in step 3) is calculated. (r) , the modulation and coding scheme used for the initial data packet and the retransmitted data packet is determined based on the current queue length 2. The method according to claim 1, characterized in that The cost function g(t) used in step 3) is designed as follows: 3.1) To meet the deterministic delay constraint, a virtual queue Q is constructed v (t), satisfying the following evolution process: Where O(t) is the delay violation, and O is the initial queue delay violation. f (t), retransmission queue delay violation O r (t) and the amount of data that failed to be retransmitted b r (t) the sum of the three items; O(t)=O f (t)+O r (t)+b r (t); Initial queue delay violation O f (t) can be calculated as follows: Where x(t) is the queue scheduling factor, and the retransmission priority strategy is adopted, that is, x(t)=1{Q r (t)>0}∈{0,1},r(m f (t)0 is the mth f (t) The amount of data transmitted when the modulation and coding scheme is used, a(t) is the amount of data arriving; Indicates the initial queue threshold corresponding to the violation delay; Retransmission queue delay violation O r (t) is calculated as follows: Where r(m r (t)0 is the mth r (t) The amount of data transmission when using the modulation and coding scheme, b f (t) is the amount of erroneous data generated in the initial transmission, Indicates the retransmission queue threshold corresponding to the violation delay; 3.2) The cost function g(t) is where t (r) is the starting point of the rth update frame, W is the weighting factor of the average delay penalty, and by setting an appropriate value, the average delay is given full attention and the deterministic delay constraint is satisfied. D(t) is the delay at time t; Where λ is the data arrival rate, α is the retransmission ratio, is the retransmission rate, α and It can be obtained through statistics.

Images