Distributed Decision Contextualization via Machine Learning-based Reverse Parametrization

This paper presents a smart-machine-based decision-support framework that facilitates contextualization of individual decisions in multi-agent operations in shared airspace usage. In specific, the presented research aims to enhance the strategic decision-making from the different smart vehicles and operators looking to use shared resources in the federated airspace management architecture. Typical challenges addressed by the presented work are the high demand for decisions in a distributed and data-centric operational environment, and the information sparsity therein due to the lack of data and intent sharing among the competitive business operators. The presented research utilizes artificial intelligence and machine learning based reverse parametrization of the decision-making factors, and, thereby, offering a viable pathway towards holistic contextualization of operational intents and conflict resolution through inter-vehicle or inter-operator negotiation for resource sharing. This paper discusses the feasibility study and the utilization of this reverse parametrization approach through a cargo delivery scenario with multiple service providers and different cargo priorities. The preliminary results presented in the paper demonstrate the comparison of different machine learning models in predicting the undisclosed parameters and their parametric weights, both for cases with and without correlations between the disclosed and undisclosed parameters. The initial findings are encouraging as they show very good conformity between the predicted and actual parameters and preferences.