Signal Overhead Reduction for AI-Assisted Conditional Handover Preparation

Abstract:
Due to complexity of handover (HO) management, artificial intelligence (AI) is envisioned as a promising candidate to assist HO procedures. However, the required signal overhead is a huge drawback of many AI-based approaches. In this paper, we focus on signal overhead reduction for AI-assisted conditional HO (AI-CHO). In this AI-CHO, a classifier performs CHO preparations. The users transmit their received measurements from serving and neighbor cells to the classifier imposing a heavy burden on network. To this end, we propose a 2-step solution which includes employing an additional simple classifier at user side to prevent transmission of unnecessary measurement reports. In addition, a pattern for number of bits compressing the remaining data with uniform scalar quantization is selected. The bit allocation is determined in accordance with the heuristic that measurements of stronger links can provide more information for the CHO classifier and need a higher quantization precision. The proposed approach results in remarkable gain in overhead reduction, i.e., 53% for our simulation setup, while providing similar outcome in terms of mobility key performance indicators such as radio link failure.