SSM视角下δ形传输拥塞的物理层解决:以UHF_RFID为例

Effective machine-type communication (MTC) is one of the key technologies for internet of things (IoT). The typical application of MTC is a delay tolerant network (DTN) connected with a large number of devices simultaneously. Usually, the connection is performed via random access methods on a media access control (MAC) layer. However, the mechanism of the random access will be to wait or retransmit if collisions happen, and thus its bottleneck is difficult to solve the δ-shape transmission congestion caused by sudden communication. This proposal will introduce a state space model (SSM) and take a ultra-high frequency radio frequency identification (UHF_RFID) system as a study case to directly recover the collision on a physical (PHY) layer. This will reduce the network congestion during the sudden communication. First, we will establish a state and a measurement equation to correctly express time-varying collision signals, add a constraint to make the equations determined and estimate the collision signal attenuation coefficient and the number of the collision tags. Second, we will separate the collision signals via sequential state estimation and leverage its differential and adaptive form to change space complexity into time complexity. Finally, we will decode the separated signals to obtain the collision tags’ IDs based on hidden Markov model (HMM). Compared with traditional PHY-layer methods, the study in this proposal is expected to circumvent the problem of frequency drift, underdetermined solution and high computation complexity to obtain a higher collision recovery efficiency. The study results in this proposal will provide a new idea to reduce the congestion of network accessed by excessive devices and a reference to the improvement of the random access for IoT.

实现有效的机型通信(MTC)是物联网(IoT)的关键技术之一，MTC典型应用是容迟网(DTN)与大批设备同时连接，该连接通常采用媒体接入控制层的随机多址实现，但其机制为冲突后重发或等待，瓶颈是难以解决突发或定期通信造成的δ形传输拥塞。本项目拟引入状态空间模型(SSM)，以超高频射频识别(UHF_RFID)为例，在物理层直接恢复冲突以减少网络拥塞。首先构建状态和观测方程表达冲突信号时变特征，并加上约束条件减少方程欠定性，同时估计衰减系数和标签数等参数；再次采用逐次状态估计分离冲突信号，并利用其差分和自适应形式将空间复杂度转化为时间复杂度；最后以隐马尔科夫链为基础，对分离信号解码以获得标签ID。与传统物理层方法相比，拟研究内容可望解决频率漂移、欠定求解和高计算复杂度等问题以得到较高的冲突恢复效率，拟研究成果将提供一种新思路去减少接入过多时的网络拥塞，为完善随机多址面向IoT的应用提供借鉴意义。

本项目的研究主要围绕SSM模型在各信息系统的应用所展开，包括SSM模型在UHF_RFID通信系统的应用，及SSM模型在其他传感器通信系统的应用，如轨枕定位通信系统、工业噪声干扰下的语音通信系统，其目标均是提高系统的通信效率，减少干扰，增强通信的可靠性。另外，本项目所完成的研究内容还包括将SSM模型应用于脑认知及脑神经信号建模中，如神经元脉冲信号、功能核磁共振图像信号的处理，解决了神经解码、神经元分类和信号建模等问题。