Akademik Veri Yönetim Sistemi
11th International Conference on Modelling and Simulation for Autonomous Systems, MESAS 2024, Chania, Yunanistan, 1 - 03 Ekim 2024, cilt.15761 LNCS, ss.68-81, (Tam Metin Bildiri)
In closed-loop simulations, scripting the missions of autonomous agents before scenario execution is challenging. This is partly because of emergent multi-agent behavior, especially for scenarios of long-duration. Moreover, the users of simulation software are usually non-programmers. This makes traditional scripting a difficult tool for programming multiple agents. Existing solutions -such as scripting languages and visual scripting- mainly stem from the game industry where popular game engines provide tools for programming agent behavior for game developers. Closed-loop simulation autonomy requirements differ in the sense that, a certain autonomy alongside a long or complicated scenario should be defined, which requires a higher-level on top of low-level agent autonomy. Reactive and inherently short-sighted agent programming tools based on state-based automata, such as behavior trees and finite state machines, can fall short for the definition of long-term plans. Moreover, these methods require some learning and adaptation effort for non-programmer users or operators. This paper presents a visual agent planning solution based on Business Process Model and Notation (BPMN). Most end-users are relatively familiar with the proposed visual notations in general. Therefore, the method provides human-readable and easy-to-comprehend conditional agent planning capability. The proposed mission programming approach guides the user to build concurrent and synchronized mission plans for implementing tactical behavior. We also discuss some details about the method and argue that the proposed approach can be promising for autonomous platform programming at the mission level.