Consider a video-conferencing system or a wireless video sensor network with cameras housed on very thin, resource-limited nodes. Here, it is not feasible to stream high definition video because of the prohibitive computational requirements of real-time video compression and streaming. Active wireless spaces enables such tasks to be handled by the embedded processing nodes. This approach can also enhance user experience in streaming video applications for mobile users, e.g., by making it possible to send high definition video with multiple camera angles at a sporting event. In this case, portable viewing devices like Smartphones or iPhones will be able to view custom combinations of resolutions and camera angles. User agents running in the embedded processing nodes can process the enhanced high-bandwidth streams from the source and send out a video stream customized for the viewer. Other possible uses include rendering of 3D graphics in real time for portable displays as part of interactive gaming or virtual reality applications.

Figure 1: Example Network for Active Wireless Spaces

Our research addresses the resource allocation and scheduling challenges that arise as part of designing such systems. One of the main problems is of allocating a large set of client nodes to a limited number of the embedded computation nodes. Such an allocation would have to take into account the constraints on both the wireless link capacity and the processing capabilities of the embedded elements, as well as satisfy the end-to-end scheduling deadlines for the applications running on the clients. The problem becomes more complex when the clients are mobile, or the access points in the AWS have bandwidth limitations. We are developing a set of fast algorithms to perform this scheduling under various conditions of the system. These algorithms have been shown to perform close to optimal exhaustive approaches, while reducing the execution times by three or more orders of magnitude.

back to top