Lifetime maximization through adaptive power allocation in reconfigurable system design for wireless systems

The recent advances in technologies including wireless communications, VLSI design, and multimedia processing, have offered much stronger processing and communication capabilities to wireless devices and networks. As a result, the rapid increase of performance and power demands by battery-powered wireless devices and systems are quickly outpacing the development of battery technologies, leading to an increasingly wider “battery gap” that limits the lifetime of wireless systems. In this thesis, we will present several lifetime maximization techniques through reconfigurable system design to bridge the battery gap. Our proposed reconfigurable VLSI architectures can obtain the optimal power-quality tradeoff by adjusting the operation points to the ever-changing environments including energy status, characteristics of input data, operation condition of other devices/building blocks, as well as the performance requirement. Our designs consider a wide range of issues from the energy management of a single wireless device to multiple wireless devices that form a wireless network. The lifetime maximization problems are formulated as different types of optimization problems. A variety of simple and efficient algorithms are proposed to find the optimal solutions adaptively. For a single battery-powered wireless device, lifetime maximization is equivalent to minimize the device’s energy consumption adapting to the environment in a most energy-efficient way. Adaptive energy allocation technique is proposed in which more energy is allocated to the processing of the content parts with the more contribution to quality. Two applications of energy-efficient wireless multimedia processing, a discrete cosine transformation (DCT) design which is widely used in image and video encoding as well as a decoding framework to decode JPEG2000 images, are implemented using the proposed adaptive energy allocation techniques. We then extend our work to the lifetime maximization of wireless systems with multiple devices. The devices are powered by separate batteries and network lifetime is defined as the time duration until the first device failure. Instead of minimizing energy consumption of every device, battery-aware design is required to adjust the operation of the devices adapting to battery energy distribution inside the wireless system. We first study a two-device wireless system and propose a simple optimal configuration allocation algorithm to efficiently control the operation of both devices adapting to their battery status. The far more complicated wireless sensor networks consisting of many more devices is then studied. We formulate the lifetime maximization problem through joint routing and sleep scheduling as a non-convex problem. We tackle this problem by transforming it into an equivalent signomial programming (SP) problem which is then solved approximately through an iterative geometric programming (IGP) algorithm. The proposed algorithm serves as a useful benchmark to evaluate practical heuristics that endeavor to maximize the network lifetime. Numerical results and comparisons with various conventional non-adaptive and adaptive systems are provided to demonstrate the potential of our proposed techniques. It is shown that the proposed designs substantially prolong the network lifetime of wireless devices and systems