A system and method for optimizing pre-departure sequencing instantiates independent agent processes for each aircraft. Each independent agent is trained via multi-agent reinforcement machine learning with an existing set of data to maximize that agent's individual reward within established bounds and while cooperating with each other agent. The interaction of the independent agents produces an optimized pre-departure sequence. The individual reward is defined with respect to a change between the current timestamp and the newly suggested timestamp. A lower delta results in a higher reward.
PRE-DEPARTURE SEQUENCING USING MULTI-AGENT REINFORCEMENT LEARNING
2024-05-23
Patent
Electronic Resource
English
PRE-DEPARTURE SEQUENCING USING MULTI-AGENT REINFORCEMENT LEARNING
European Patent Office | 2024
|