Predictive HARQ Feedback Using ML for Low-Latency Retransmission

The patent applies preliminary action by performing channel decoding before complete data reception. The UE decodes the received portion of downlink data and generates HARQ feedback (ACK/NACK) based on this partial decoding, rather than waiting for the entire data packet to be received and decoded. This enables early feedback transmission, reducing latency while maintaining sufficient reliability through the decoding of available data portions.

The patent applies self-service by enabling the UE to autonomously perform machine learning-based predictive feedback generation without requiring base station intervention in the decision-making process. The UE's machine learning model independently analyzes channel conditions and data characteristics to predict whether retransmission is needed, generating feedback automatically based on local observations and learned patterns, thereby reducing overall system complexity despite the advanced processing required.

The patent applies parameter changes by adapting the machine learning model's input parameters and decision thresholds based on varying channel conditions and data characteristics. The model dynamically adjusts its prediction criteria according to factors such as signal strength, interference levels, and data encoding parameters, allowing flexible feedback timing while maintaining prediction accuracy through adaptive parameter adjustment rather than relying on fixed precision measurements.