Solving Complex Problems With Computational and Interfacing Tools
Denis Lalanne · 1997
this article is concerned with the reallocation of resources to the scheduled tasks. The resource allocation problem in reactive scheduling is a hard one to solve. It is often bound to external, to the schedule, events that are hard to be anticipated in advance by the system. Even if the rescheduler succeeds in taking into account all the needed parameters it may end up with an underconstrained problem that produces a large number of solutions, thus converting the initial combinational problem to a discrete optimization one. In both cases the active involvement of an expert human operator is essential. The human expertise, knowledge and reasoning have to be incorporated into the problem solving process. There are already attempts to blend interactivity with artificial intelligence techniques in order to solve scheduling problems [2]. The user is actively involved into the process of problem solving and aids into the performance of the underlying computational engine. He either introduces additional parameters, constraints and rules or he controls the performance of the supplied computational tools. But this is only the one direction from the bi-directional communication link between the human intelligence and the machine's computational engine. In order to succeed in an effective interaction model the machine has to respond in a meaningful way that will help the user's understanding and reasoning process. For example, in the case of an underconstrainted problem in rescheduling, the system may return a set of possible solutions. Each of the solutions, if applied to the schedule, returns the schedule into a new equilibrium. But which of the solutions is the optimal one? How can the user decide? He has to go through all of them, examine them, reason about their applicabi...