Page 27 - ATC Special Bulletin Series - Training & Simulation 2022-01
P. 27

| 01 | TRAINING & SIMULATION | did not explicitly extend to deployment phases (V4-V6). AI based solutions will require validation through the deployment process and into operations monitoring as a means of scrutinising whether algorithmic models are continuing to perform to expectation, delivering anticipated benefits and are learning as expected. While there remain uncertainties associated with extending the validation process through into live operations, an assurance- based view may be more beneficial. It has never been expected that the human components of ATM systems be correct 100% of the time. Instead, potential errors or unusual events are identified and mitigated against to an acceptable level ahead of deployment. In this vein, the same may be applied to new system elements that are intended to replicate human tasks. All of this begs the question of which validation techniques will be effective in assessing complex automation where human actors are also involved. As AI-based technologies are introduced to ATM, how will this influence the role of the Human-In-The Loop (HITL) Real-Time Simulations (RTS) as the gold-standard of validation? It may be that predictive modelling and scenario- based simulations become more prevalent in validating the aspects of the solution that carry greatest potential risk. This is likely to be more scalable than a typical RTS-approach, in which each exercise may take months to plan, prepare, execute and report on before a design can be iterated on. 2. New Generation Planning and Co-ordination With the increase in the availability of data and the improved accuracy and timeliness of that data, researchers are improving the ability to predict, co-ordinate and optimise systems involving several stakeholders. It is anticipated that such projects will impact Airport Collaborative Decision Making (A-CDM), fleet management, regional flow / network management and in the long term may even open the door to global network management solutions. Such broad concepts will be full of interdependencies, and challenges are presented in terms of how the current E-OCVM could be properly applied to remove risk from such deployments. In years past, the validation process could be expected to encapsulate a single function, procedure or infrastructure change; perhaps only impacting operations on a local ANSP or airport level. The planned future of such solutions involve interactions between a list of highly-complex systems and operational layers. New thinking is required to establish a methodology that can scale up from a more traditional ATM concept to one that supports a more systematic perspective. Indeed, the fundamental idea of ‘deployment’, where the old element is removed and the new one is ‘switched-on’ may become outdated. Where transitions from legacy to solution 27 | 

   25   26   27   28   29