Predictive, Sensor-Assisted Wireless Multimedia Systems

美国康奈尔大学预测、传感器辅助无线多媒体系统未来的PCS系统将为移动用户提供多种服务，包括语音、高速数据、视频、电子邮件、电话会议和医学图像的传输。智能PCS网络将被要求提供最低限度的呼叫阻塞的信道接入，协商服务质量，分配资源，并在用户的整个会话期间跟踪用户。建议开发一系列专家系统来优化随机接入、资源分配和移动性管理协议。这些专家系统将使用神经网络、人工智能和知识表示领域的许多技术和技巧。在提出的智能PCS网络中，有三级代理：移动用户、基站控制器(BSC)和移动交换中心(MSC)。假设PCS系统是蜂窝的，多个BSC与单个MSC相互作用。建议确定一种表示和利用每种类型的代理可用的信息的方法。在物理层，带内传输功率水平将由低成本传感器(辐射计)组成的网格进行跟踪。传感器将根据时间和位置提供功率测量。在该方案中，这些信息可以用于基于BSC的智能ALOHA多址接入方案的开发。描述了神经网络，该神经网络可以有效地使用传感器数据来估计在各个碰撞事件中涉及的用户数量。该信息又被用来估计积压用户的数量并选择最佳退避算法，从而最大化ALOHA信道吞吐量。BSC和MSC也可以使用传感器网格来支持资源分配、切换和跟踪调度的用户传输。在拟议的研究计划中，将为PCS操作的三个不同阶段开发基于传感器的专家系统：系统规划和布局、多址接入和自适应资源分配。更高级别的信息可以在BSC和MSC处获取和使用。BSC将获得关于作为时间函数的本地用户的带宽和服务质量要求的信息。BSC还将跟踪本地信道条件，并为小区开发详细的传播模型。该模型将被用作电池内电源控制的辅助工具。预计BSC将在本地资源分配和功率控制方面发挥自治代理的作用。MSC将负责在小区之间分配资源。将开发分配协议以响应于由BSC获取并传递给MSC的交通信息在小区之间移动资源。BSC还将传播和用户跟踪信息传递给MSC。MSC将使用该信息来创建支持切换决策过程的全局传播和用户跟踪模型。将努力为个人用户编制使用倾向简档。收集的用户资料将用于开发全球资源分配模型。

ABSTRACT NCR-9725251 Stephen Wicker, Terrence Fine, and Joseph Halpern Cornell University Predictive, Sensor-Assisted Wireless Multimedia Systems The PCS systems of the future will provide a variety of services to mobile users, including voice, high speed data, video, e-mail, teleconferencing, and the transfer of medical Images. The intelligent PCS network will be required to provide channel access with a minimum of call blocking, negotiate quality of service, allocate resources, and track users throughout their sessions. It is proposed to develop a series of expert systems that will optimize random access, resource allocation, and mobility management protocols. These expert systems will use a number of technologies and techniques from the fields of neural networks, artificial intelligence, and knowledge representation. There are three levels of agents in the proposed intelligent PCS network: mobile user, Base Station Controller (BSC), and the Mobile Switching Center (MSC). It is assumed that the PCS system is cellular, with several BSC's interacting with a single MSC. It is proposed to determine a means for representing and exploiting the information available to each type of agent. At the physical layer, in-band transmitted power levels will be tracked by a grid of low-cost sensors (radiometers). The sensors will provide power measurements as a function of time and location. In the proposal it is shown that this information can be used in the development of a BSC-based intelligent ALOHA multiple access scheme. Neural networks are described that can efficiently use the sensor data to estimate the number of users involved in individual collision events. This information is used in turn to estimate the number of backlogged users and to select an optimal backoff algorithm, thus maximizing Aloha channel throughput. The sensor grid can also be used by the BSC and MSC to support resource allocation, handoffs, and the tracking of scheduled user transmissions. In the proposed research program, sensor-based expert systems will be developed for three distinct phases of PCS operation: system planning and layout, multiple access, and adaptive resource allocation. Higher level information can be acquired and used at the BSC's and the MSC. The BSC's will obtain information regarding the local users' bandwidth and quality of service requirements as a function of time. The BSC's will also track local channel conditions, and develop a detailed propagation model for the cell. The model will be used to as an aid to power control within the cell. The BSC is expected to act as an autonomous agent in local resource allocation and power control. The MSC will be responsible for allocating resources between cells. An allocation protocol will be developed to move resources between cells in response to traffic information acquired by the BSC's and passed to the MSC. The BSC's will also pass propagation and user tracking information to the MSC. The MSC will use this information to create a global propagation and user tracking model that will support the handoff decision process. An effort will be made to develop usage tendency profiles for individual users. The collection of user profiles will be used to develop models for global resource allocation.