Texture-Based Heuristics for Scheduling Revisited

tics with the same consistency techniques (Nuijten, 1994). In this paper we re-evaluate texture-based heuristics in light of recent advances in scheduling technology and show that on two job shop scheduling problem sets (a widely used set of Operations Research benchmark problems and a set of randomly generated, hard problems) a texture-based heuristic outperforms heuristic commitment techniques found in the literature. The n × m Job Shop Scheduling Problem Given are a set of n jobs, each composed of m totally ordered activities, and m resources. Each activity Ai requires exclusive use of a single resource Rj for some processing duration duri.. There are two types of constraints in this problem: • Precedence constraints between two activities in the same job stating that if activity A is before activity B in the total order then A must execute before B (that is, A → B). • Resource constraints specifying that no activities requiring the same resource may execute at the same time. Jobs have release dates (the time after which the activities in the job may be executed) and due dates (the time by which the last activity in the job must finish). In the decision problem, the release date of each job is 0 and a global due date is D. The problem is to determine whether there is an assignment of a start-time to each activity such that the constraints are satisfied and the maximum finish-time of all jobs is less than or equal to D. This problem is NP-complete (Garey and Johnson, 1979).