August 23, 2019 –
Title: Dynamic Resource Management in Lossy Real-Time Wireless Networks
Ph.D. Candidate: Tao Gong
Major Advisor: Dr. Song Han
Associate Advisors: Dr. Bing Wang, Dr. Shengli Zhou
Day/Time: Friday, Aug 23rd, 2019 10:00AM
Location: ITE 119
Along with the rapid development and deployment of real-time wireless network (RTWN) technologies in a wide range of time- and safety-critical industrial applications, effective packet scheduling algorithms have been playing a critical role in RTWNs to achieve desired Quality of Service (QoS) for real-time sensing and control, especially in the presence of unexpected disturbances. Most existing solutions in the literature focus on either static or dynamic communication schedule construction to meet the desired QoS requirements but have common assumptions that all wireless links are reliable, and the network topologies are prior known. Although these assumptions simplify the algorithm design and analysis, they are not realistic in real-life settings where wireless links are not reliable, and the network topologies are subject to change in the runtime.
To relax the assumption on the perfect network links in existing work, in the first part of this dissertation, we propose RD-PaS, a reliable dynamic packet scheduling framework for RTWNs, to guarantee the QoS in lossy wireless environments. RD-PaS can not only construct static schedules to meet both the timing and reliability requirements of end-to-end packet transmissions in RTWNs for a given periodic network traffic pattern, but also construct new schedules to handle abruptly increased network traffic induced by unexpected disturbances while minimizing the impact on existing network flows. Given that RD-PaS relies on a central controller to collect disturbance information and make schedule change decisions, which is not scalable in large-scale RTWNs, we further propose FD-PaS in the second part of this dissertation to support distributed schedule change which leads to a much faster response time than RD-PaS and does not require a central controller. FD-PaS allows the involved nodes to make consistent schedule change decisions immediately upon the detection of the disturbances. MP-MAC, a novel MAC layer design compatible with the existing 802.15.4 standard is further proposed to allow multi-level over-the-air packet preemption. Both MP-MAC and the FD-PaS packet scheduling framework have been implemented on a RTWN testbed for design validation and performance evaluation.
In the third part of this dissertation, we focus on improving the QoS in large-scale RTWNs in which the network topologies could be dynamic in the runtime. RD-PaS and FD-PaS cannot be applied on these networks as they require static network topology. A randomized scheduler has been implemented in such networks to construct baseline communication schedules. While such schedules allow networks to be operational, they suffer from high latency for end-to-end packet transmissions. To optimize the transmission latency, we propose a partition-based scheduler to reserve network resources by partitions, where those reserved resources will be utilized during topology changes to minimize the impact to the network latency. Given that the resource reservation is based on estimation, it may not be accurate in the runtime. In such case, the latency may be increased due to the fail-safe mechanism of insufficient resource reservation. To address this problem, we propose a dynamic partition adjustment algorithm to compensate for those inaccurate reservations and thus provide bounded QoS performance in the runtime. The partition-based scheduler has been implemented in our large-scale RTWN testbed for design validation and performance evaluation.