| In the modern world, where users send requests via the Internet requesting tasks to be performed, efficient scheduling of those requests becomes a major concern. This thesis concentrates on a robotic domain. The robot is working in an unpredictable environment where in, the system must be able to handle requests arriving at a continuous rate and provide an optimal schedule in real time.; In this thesis, we describe a research project, which performs optimization of robotic task schedules using genetic algorithms. To estimate the task execution times, a set of training examples is collected by running a simulator. A neural network is then trained using these collected examples. The output of the neural network is fed into the genetic algorithm allowing it to get an estimate of the total time required for the schedule. We describe the scheduling system and provide experimental results demonstrating the effectiveness of this research. |
