Real-Time Adaptive Resource Management Middleware For Next Generation Avionics And Algorithms

Avionics is developing toward micromation, integration and intelligentization. New generation avionics is a DRE (Distributed Real-time Embedded), which integrates the wireless networks with the wired networks and has explicit collaboration and a mass of information exchange. Avionics software becomes increasingly complicated, with the integration trend. Traditional real-time middleware faces new challenges: 1) Traditional real-time middleware does not support dynamic and unpredictable workload management. 2) Traditional real-time middleware does not support real time transportability although it supports function transportability on different platform. 3) Traditional real-time middleware does not separate the function implementation and the QoS (Quality of Service) guarantee, reducing the component reusability greatly.In order to adapt to the trend of avionics integration, this dissertation proposed a RTARMM (Real-Time Adaptive Resource Management Middleware) for the new generation avionics and three adaptive algorithms. The objectives are to support the unpredictable workload management, and to support real time transportability to settle the austere problems raised in the avionics update and maintenance. The contributions of this dissertation are as follows.1) Analyzed the trend of avionics and introduced the architecture of America F-22 and the blue print of Chinese DCS (Distributed Computer System) of the forth generation flighters. Then according to the actual situation of Chinese avionics development, proposed the design idea, architecture and implementation scheme of the RTARMM. Next introduced the QoS to the real-time tasks, providing the support for the QoS mechanism from the software architecture, and designed the special components and modules to implement multifold QoS mechanisms. Moreover, realized the objective of separating the function implementation from the QoS guarantee.2) Proposed an optimal reward based reallocation algorithm for the periodic tasks, which applies the idea of two-level scheduling. A top-level scheduling dynamically adjusts the tasks' QoS on the case of workload variation to maximize the system performance and ensure the system schedulability. A lower-level algorithm, such as RM (Rate Monotonic), is responsible for the actual task scheduling.3) Proposed a performance control algorithm for the DRE systems that have the end-to-end task model, which features a distributed feedback loop that relates the per-task metric (meeting individual end-to-end deadlines) to the aggregate metric (utilization). This algorithm considers aperiodic tasks and regards both task arrivals and execution time variation as the extemal factors that cause the system unpredictability. It integrates admission control with classical feedback control, which is able to dynamically determine the QoS of incoming tasks and guarantee the end-to-end deadlines of accepted tasks, at the same time improve the system throughput.4) Proposed a utilization control algorithm for the DRE systems that have the end-to-end task model. This algorithm integrates the admission control with the optimization control theory, maps the system into a muilti-input-multi-output control problem, and adopts the state space to describe the system behavior. Next proposed a linear quadratic optimal control model to realize the utilization control for the DRE systems, which is capable of providing better QoS guarantee.The research on the real-time adaptive resource management middleware and algorithms for the new generation avionics will provide the theory and technique supports for the software development of the new generation avionics in our country.