Allocating and Executing Tasks Sample Clauses

Allocating and Executing Tasks. ‌ Moving beyond the structure of each of the elements of our model as presented in Chapter 3, we are able now to consider, in relation to our model, the behaviour of agents so that tasks are completed, either by a single agent executing them in full, or through multiple agents executing distinct parts of a task. Here, agents are considered to be heterogeneous, with each agent offering a different set of services. However, the core processes that each agent performs are the same. All agents continually perform two processes: they analyse received tasks to determine whether they can be executed, and if possible execute them; and they reallocate any tasks that cannot be executed, to other agents. We consider each of these separately. As stated earlier, in considering the management and execution of tasks, we adopt a base model similar to that of others (e.g. [70, 2]). For simplicity, we assume that time is discrete, with a number of time steps, in each of which an agent a must: manage its tasks list t; and execute tasks in its running tasks set rt. In managing tasks, a first places any received tasks in the list t. Then each task t ∈ t is reviewed: if t ’s initial requirement can be satisfied directly by a (because a offers the required service and has the capacity to do so), a removes t from t, and adds the tuple (t, si) to rt where si is the service instance executing t; if a can satisfy t ’s initial requirement, but does not have capacity to do so, then the task remains on the list, waiting for capacity to become available; finally, if a cannot satisfy t ’s initial requirement, then it must allocate the task to another agent. After t is updated, execution begins on an agent’s running tasks list, rt, by satisfying each task’s initial requirement. Since tasks require a service for a specified time (in time steps), srt, they remain in rt until srt has elapsed, at which point they are removed. At this point, if the task’s initial requirement rx is also the task’s only requirement, then the task is complete, otherwise a subtask is created for every branch from rx in the requirement ordering tree. The initial requirement of each subtask is r1, r2, . . . , rn, such that there are n subtasks, and (rx, r1), (rx, r2), . . . , (rx, rn) ∈ rot. Each subtask is added to t so that it can subsequently be executed or allocated. Through the repeated use of this process, each task an agent receives will be managed, and either executed or reallocated to another agent. Finally...