Comparing Strategies for Agent-Based Dynamic Scheduling on the MADySUM Workbench

Static systems that aim at scheduling tasks to achieve load-balancing, fail to provide a simple solution for dynamic system behavior. The necessity to reschedule, nonadaptability and the strategy complexity used by many approaches to scheduling are drawbacks easily overcome by an agent-based dynamic scheduling system. One of such systems is the MADySUM (Multiagent Dynamic Scheduling at the University of Minho) one, presenting the user with a simple interface, high-scalability and flexibility. The MADySUM workbench provides a set of agent strategies: threshold, k*y < limit, less sent, most sent and pheromone trail. Testing reveals that, through MADySUM, less sent is the best strategy for orders composed of equal task times and pheromone trail is the best for the different execution ones.