| CPC G06Q 10/063118 (2013.01) [G06F 16/9024 (2019.01); G06N 3/126 (2013.01); G06Q 10/06313 (2013.01); G06Q 10/06316 (2013.01); G06Q 10/063114 (2013.01); G06Q 10/063116 (2013.01)] | 16 Claims |
|
1. A computer-implemented method comprising:
obtaining a graph representing an environment in which multiple agents are to act, the graph including nodes relating to tasks and edges relating to traversable paths between the tasks, a schedule for the multiple agents being determined to perform a set of the tasks;
determining the schedule for performing the set of the tasks, the schedule including, for each agent, a sub-schedule including tasks to be performed and, for each task to be performed, a corresponding scheduled range of times over which to perform the task;
evaluating, for the schedule, a probability of completion of the schedule, the probability of completion of the schedule being the probability that each task in the set of the tasks will be completed subject to the condition that each agent has a predefined failure rate, wherein evaluating the probability of completion includes:
determining a completion region within state space, wherein the completion region is a range of potential states of the environment that could be achieved by following the schedule such that every task has been visited at least once by any of the agents;
identifying one or more sub-regions within the completion region, each sub-region relating to a given combination of tasks being performed simultaneously by the agents according to the schedule; and
determining a total probability mass within the completion region including determining a sum over all sub-regions within the completion region of the probability of completion of each sub-region, wherein the probability of completion of each sub-region is determined by determining an integral of the probability of completion over the sub-region; and
determining an updated schedule based on an objective function that aims to provide a greater probability of completion of the schedule; and
sending the updated schedule to the agents and controlling at least one autonomous system by implementation of the updated schedule.
|