Resource Management in E-health Systems
E-health systems are the information and communication systems deployed to improve quality and efficiency of public health services. Within E-health systems, wearable sensors are deployed to monitor physiology information not only in hospitals, but also in our daily lives under all types of activities; wireless body area networks (WBANs) are adopted to transmit physiology information to smartphones; and cloud servers are utilized for timely diagnose and disease treatment. The integrated services provided by E-health systems could be more convenient, reliable, patient centric and bring more economic healthcare services. Despite of many benefits, e-health systems face challenges among which resource management is the most important one as wearable sensors are energy and computing capability limited, and medical information has stringent quality of service (QoS) requirements in terms of delay and reliability. This thesis presents resource management mechanisms, including transmission power allocation schemes for wearable sensors, Medium Access Control (MAC) for WBANs, and resource sharing schemes among cloud networks, that can efficiently exploit the limited resources to achieve satisfactory QoS.
First, we address how wearable sensors could energy efficiently transmit medical information with stringent QoS requirements to a smart phone. We first investigate how to provide worst-case delay provisioning for vital physiology information. Sleep scheduling and opportunistic channel access are exploited to reduce energy consumption in idle listening and increase energy efficiency. Considering dynamic programming suffers from curse of dimensionality, Lyapunov optimization formulation is established to derive a low complexity two-step transmission power allocation algorithm. We analyze the conditions under which the proposed algorithm could guarantee worst-case delay. We then investigate the impacts of peak power constraint and statistical QoS provisioning. An optimal transmission power allocation scheme under a peak power constraint is derived, and followed by an efficient calculation method. Applying duality gap analysis, we characterize the upper bound of the extra average transmission power incurred due a peak power constraint. We demonstrate that when the peak power constraint is stringent, the proposed constant power scheme is suitable for wearable sensors for its performance is close to optimal. Further, we show that the peak power constraint is the bottleneck for wearable sensors to provide stringent statistical QoS provisioning.
Second, WBANs can provide low-cost and timely healthcare services and are expected to be widely adopted in hospitals. We develop a centralized MAC layer resource management scheme for WBANs, with a focus on inter-WBAN interference mitigation and sensor power consumption reduction. Based on the channel state and buffer state information reported by smart phones deployed in each WBAN, channel access allocation is performed by a central controller to maximize the network throughput. Note that sensors have insufficient energy and computing capability to timely provide all the necessary information for channel resource management, which deteriorates the network performance. We exploit the temporal correlation of body area channel such that channel state reports from sensors are minimized. We then formulate the MAC design problem as a partially observable optimization problem and develop a myopic policy accordingly.
Third, cloud computing is expected to meet the rising computing demands. Both private clouds, which aim at patients in their regions, and public clouds, which serve general public, are adopted. Reliability control and QoS provisioning are the core issues of private clouds and public clouds, respectively. A framework, which exploits the abundant resource of private clouds in time domain, to enable cooperation among private clouds and public clouds, is proposed. Considering the cost of service failure in e-health system, the first time failure probability is adopted as reliability measures for private clouds. An algorithm is proposed to minimize the failure probability, and is proven to be optimal. Then, we propose an e-health monitoring system with minimum service delay and privacy preservation by exploiting geo-distributed clouds. In the system, the resource management scheme enables the distributed cloud servers to cooperatively assign the servers to the requested users under a load balance condition. Thus, the service delay for users is minimized. In addition, a traffic shaping algorithm is proposed, which converts the user health data traffic to the non-health data traffic such that the capability of traffic analysis attacks is largely reduced.
In summary, we believe the research results developed in this dissertation can provide insights for efficient transmission power allocation for wearable sensor, can offer practical MAC layer solutions for WBANs in hospital environment, and can improve the QoS provisioning provided by cloud networks in e-health systems.