CQI Prediction Through Recurrent Neural Network for UAV Control Information Exchange Under URLLC Regime

Bartoli, Giulio; Marabissi, Dania

doi:10.1109/tvt.2022.3152408

Cited by 11 publications

(5 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another CNN, inspired by the image superresolution technique was proposed by [22]. In [9], [17], [23]- [29] a channel state information predictor is proposed based on a recurrent neural network and its modifications, such as Long-Short Term Memory (LSTM) [30], [31] network. The simulations and experimental studies confirm that the neural network can pick up and learn the channel evolution patterns in various fading channels, including the Rayleigh channel [24], as well as experimental measurements [32].…”

Section: Related Workmentioning

confidence: 99%

“…Furthermore, the data for evaluating and testing the algorithms need to be considered while designing the channel prediction algorithms. As the NN algorithms are data-driven, the best option for closest-to-life data needs to be experimentally obtained, which is rarely used, and channel modeling data is often used instead [23].…”

Section: Related Workmentioning

confidence: 99%

“…where l is the weighted error of the prediction with different weights for overestimation and underestimation, q is the quantile we aim to predict, y i is the element of the true value vector, and ŷi is the element of the neural network output vector, and the size of truth values and predictions vector is N . Unlike previous works [16], [23], which also use LSTM for channel quality prediction, we can explicitly optimize for the necessary target reliability during algorithm training and reduce the chance of overestimating the channel quality. Moreover, the proposed approach can be extended to multiple quantile output values to target multiple reliability requirements with a negligible additional incurred cost.…”

Section: B Asymmetric Loss During Trainingmentioning

confidence: 99%

“…However, during CQI quantization some channel information is lost, as the continuous SNR values are converted to values of 0 to 15. Similarly to [1], [23], we use the following approach to scheduling and MCS selection based on given predictions.…”

Section: Scheduling and Mcs Selection Stepmentioning

confidence: 99%

“…• The proposed algorithm, ALPACA, with asymmetric loss parameter q of 0.1, 0.2, 0.5. We note, that the quantile regression with the quantile parameter of 0.5 is equal to the mean absolute error (MAE) loss, which is symmetric, and forecasts the median, making this approach similar to the LSTM-based approaches described in papers [16], [23], hence we name the algorithm with q = 0.5 as "CNN-LSTM". The MCS is selected according to the CQI prediction to provide the 10 −5 error rate.…”

Section: B Simulation Scenariomentioning

confidence: 99%

See 4 more Smart Citations

ALPACA: An Asymmetric Loss Prediction Algorithm for Channel Adaptation Based on a Convolutional-Recurrent Neural Network in URLLC Systems

Glinskiy,

Kureev,

Khorov

2024

IEEE Access

View full text Add to dashboard Cite

A key feature of 5G systems is the Ultra-Reliable Low-Latency Communication (URLLC), which can be used for remote surgery, smart grids, industrial control, etc. URLLC requires millisecondlevel delays and very high reliability, i.e., less than 10 −5 packet loss probability. The ability to satisfy these very strict quality of service requirements depends on selecting the Modulation and Coding Schemes (MCS) for data transmissions. On the one hand, the selected MCS shall be robust enough to avoid multiple retransmissions within a small delay budget. On the other hand, the MCS shall be high-rate to reduce channel resource consumption and, thus, shall increase the system capacity for URLLC. The MCS selection problem is extremely challenging to capture the quickly varying wireless channel effects, e.g., in highly mobile scenarios, because the decision shall be made long before the actual transmission occurs. The paper proposes a novel MCS selection algorithm called ALPACA (Asymmetric Loss Prediction Algorithm for Channel Adaptation), which relies on a widely used class of convolutional-recurrent neural networks. However, in contrast to existing approaches, ALPACA explicitly considers the asymmetric error cost for channel prediction by utilizing quantile regression loss. Both real-life channel measurements and 3GPP channel models are used to evaluate the performance of ALPACA. Numerical results demonstrate the increase in the reliability and reduction in resource consumption compared with the existing MCS selection algorithms, which results in 40% growth of the network capacity.

show abstract