Gene library for real-time monitoring of large scale time-sensitive systems

For time-sensitive applications with hard real-time constraint, the utility of a decision goes to zero if the deadline is missed thus it is very important to use methodologies that can deliver solutions within their time limit. For large scale monitoring and prediction systems with small time periods this problem renders conservative optimization techniques to be useless especially because of the time they take to calculate optimal values. In order to make real-time decisions, we need to introduce methods that are computationally light and can still maintain accuracy that is close to results given by optimization methods. A gene library was formulated that stored (i) the regulatory proteins in order to select the relevant features that determined the system behavior and (ii) computationally simple mappings that mapped these relevant features to the desired system state. The proposed method was implemented in an Intelligent Transportation System scenario to determine the rollover risk. A gene library was created that eliminated the need to perform heavy computations (solving second order differential equation) while still maintaining the accuracy of prediction to +/- 4% of the actual value in the normal operating range.