In many agent-based applications, individual agents sign up to perform tasks in the future. Each agent must manage its own calendar of commitments to guide it in making additional engagements. Conventionally, an agent books each commitment to a fixed time window, but this representation gives the agent little flexibility in rearranging its commitments to accommodate new opportunities and improve its overall performance (for example, by performing similar tasks in close succession).
NewVectors’ Density-Based Emergent Scheduling Kernel (DESK) applies the concept of least commitment scheduling to the agent environment. Using this approach, a resource does not initially schedule a job for a fixed time period. Instead, a resource makes a looser commitment on the start and stop time of a job while still maintaining a full commitment to complete the job, retaining greater flexibility in managing change. Current schedulers define a fixed window where the job will execute. Thus, a lathe might be committed to start a three-hour job at 1:00 P.M. and end it at 4:00 P.M. With DESK, the resource might commit to starting the job sometime after 12:00 P.M. and completing it sometime before 5:00 P.M. Since the lathe has five hours of capacity during the window and is committing only three of them to this job, it need only commit 60 percent of its capacity to the job.