AI Model Transmission Across Link Changes Without Full Retransmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The transmission of large AI models over radio resource control (RRC) signaling is inefficient due to the need for segmentation and retransmission upon link changes or cell handovers, leading to significant air interface signaling overhead.
Innovation Solution
A method and apparatus for continuing AI model transmission across link changes by using model transmission associated information to identify untransmitted portions and either transmit them in a new access link or switch to a new model, reducing the need for complete retransmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Volume of moving object
If the AI model is divided into multiple segments and transmitted multiple times, then the transmission of large AI models becomes feasible, but the risk of incomplete transmission increases when link changes or cell handovers occur
Solution Approach 1:
The patent applies preliminary action by having the terminal store the transmitted portion of the AI model locally before link changes occur. When a link change or cell handover happens, the terminal already has the previously received segments stored, allowing it to continue receiving and merging segments without losing progress. This pre-storage mechanism ensures that even if transmission is interrupted, the terminal can resume from where it left off rather than starting over.
Solution Approach 2:
The patent implements feedback through the merging procedure where the terminal continuously monitors and combines received segments. The terminal tracks which segments have been received and successfully merges them into the complete AI model. This feedback mechanism allows the system to detect incomplete transmissions and trigger retransmission requests for missing segments, ensuring complete and reliable model delivery despite link changes.
2Adaptability or versatility
If the receiving end discards all segmented contents when a link changes or cell handover happens, then the receiving end can receive new message segments in the new link, but the transmission efficiency decreases due to complete retransmission
Solution Approach 1:
The patent applies preliminary action by having the terminal store the transmitted portion of the AI model locally before link changes occur. When a link change or cell handover happens, the terminal already has the previously received segments stored, allowing it to continue receiving and merging segments without losing progress. This pre-storage mechanism ensures that even if transmission is interrupted, the terminal can resume from where it left off rather than starting over.
Solution Approach 2:
The patent implements continuity of useful action by enabling the terminal to maintain AI model transmission progress across link changes. Instead of discarding received segments when switching links, the terminal preserves them and continues the merging process in the new link. This ensures that the useful action of model transmission continues uninterrupted, avoiding the need to retransmit already received segments and significantly improving transmission efficiency.
3Reliability
If the entire AI model is retransmitted every time a link changes or cell handover happens, then the transmission can be completed in the new link, but the air interface signaling overhead increases significantly
Solution Approach 1:
The patent applies the extraction principle by separating the AI model transmission into distinct segments and tracking which portions have been successfully transmitted. When a link change occurs, the system extracts only the missing segments for retransmission rather than sending the entire model again. This selective retransmission of only the necessary portions significantly reduces air interface signaling overhead while ensuring complete model delivery.
Solution Approach 2:
The patent implements discarding and recovering by having the terminal discard incomplete or duplicate segments and recover only the missing portions. When link changes occur, the terminal identifies which segments are missing and requests only those specific segments for retransmission. This approach avoids wasting air interface resources on retransmitting already received segments while ensuring the complete AI model is ultimately reconstructed.
Data Source
Figure 1~3
Figure 4
Figure 5
AI summary
The embodiments of the present application belong to the technical field of communications. Provided are a model transmission method and apparatus. The method comprises: acquiring model-transmission-related information, which is used for indicating that there is an untransmitted portion of a first model, wherein a transmitted portion of the first model has been transmitted on an original access link; and executing, on a new access link, the transmission of the untransmitted portion of the first model or the transmission of a second model, wherein the new access link is a link between a terminal and a first node, the original access link is a link between the terminal and a second node, and the second model is a model required for the new access link.