Adaptive Genetic Algorithm : Scheduling Hard Real - time Control

In hard real-time systems, a timing fault may lead the environment to catastrophe. Thus, a real-time program must be executed in accordance with the timing constraints speciied in the control program. Dynamic scheduling approach which has been dominant for real-time systems are unacceptable due to run-time scheduling overhead and unpredictable risks. In contrast, a static approach takes advantage of thorough search for scheduling, no scheduling overhead, and guaranteed execution of compiled program. However, nding an optimal schedule for real-time tasks with precedence is known as NP-hard problem, demanding a heuristic approach. In this paper, we propose an adaptive genetic approach to nd a valid schedule. Adaptive genetic approach implies that the algorithm relies on not only stochastic behavior of the search but also deterministic behavior based on program structure. This paper includes new computation model with arbitrary before and after constraints, description of modiied genetic algorithm, and the experimental results. This work is an endeavor to build a scheduler that is a part of the compiler for hard real-time systems (CHaRTS). CHaRTS accepts a hard real-time program and a system connguration le, then generates a sequence of executable code for target machine.