Dynamic Scripting with Team Coordination in Air Combat Simulation

Toubman, Armon; Roessingh, Jan Joris; Spronck, Pieter; Plaat, Aske; Herik, Jaap van den

doi:10.1007/978-3-319-07455-9_46

Cited by 17 publications

(8 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Over the years, there have been several attempts at using machine learning techniques for building behavior models for computer generated forces. Some approaches that have been studied are evolutionary algorithms [40][41][42], neural networks [43][44][45][46] and dynamic scripting [47][48][49], a technique originally developed for computer games. Still, the studied techniques have not been mature enough to include in commercial CGF software [9].…”

Section: Related Workmentioning

confidence: 99%

Multi-Agent Multi-Objective Deep Reinforcement Learning for Efficient and Effective Pilot Training

Källström

Heintz

2019

Proceedings of the 10th Aerospace Technology Congress, October 8-9, 2019, Stockholm, Sweden

View full text Add to dashboard Cite

The tactical systems and operational environment of modern fighter aircraft are becoming increasingly complex. Creating a realistic and relevant environment for pilot training using only live aircraft is difficult, impractical and highly expensive. The Live, Virtual and Constructive (LVC) simulation paradigm aims to address this challenge. LVC simulation means linking real aircraft, ground-based systems and soldiers (Live), manned simulators (Virtual) and computer controlled synthetic entities (Constructive). Constructive simulation enables realization of complex scenarios with a large number of autonomous friendly, hostile and neutral entities, which interact with each other as well as manned simulators and real systems. This reduces the need for personnel to act as role-players through operation of e.g. live or virtual aircraft, thus lowering the cost of training. Constructive simulation also makes it possible to improve the availability of training by embedding simulation capabilities in live aircraft, making it possible to train anywhere, anytime. In this paper we discuss how machine learning techniques can be used to automate the process of constructing advanced, adaptive behavior models for constructive simulations, to improve the autonomy of future training systems. We conduct a number of initial experiments, and show that reinforcement learning, in particular multi-agent and multi-objective deep reinforcement learning, allows synthetic pilots to learn to cooperate and prioritize among conflicting objectives in air combat scenarios. Though the results are promising, we conclude that further algorithm development is necessary to fully master the complex domain of air combat simulation.

show abstract

Section: Related Workmentioning

confidence: 99%

Multi-Agent Multi-Objective Deep Reinforcement Learning for Efficient and Effective Pilot Training

Källström

Heintz

2019

Proceedings of the 10th Aerospace Technology Congress, October 8-9, 2019, Stockholm, Sweden

View full text Add to dashboard Cite

show abstract

“…The most often-used tactics representations are rule-set, decision tree, and finite state machine (FSM). These are criticized for their defects in practice 10,11,12 , such as the inability to scale up in the complex domain, the difficulties for SMEs to understand and reuse tactic scripts, and the tedious scripting.…”

Section: Related Workmentioning

confidence: 99%

Tactics Exploration Framework based on Genetic Programming

Yao¹,

Wang²,

Li³

et al. 2017

IJCIS

View full text Add to dashboard Cite

Engagement-level simulation is a quantitative way to evaluate the effectiveness of weapon systems before construction and acquisition, minimizing the risk of investment. Though contractors have built simulation systems with high fidelity models of weapon systems and battlefields, developing competent tactics to give full play to new weapon systems in simulation experiments is labor intensive, as most classical tactics tend to be out of date. In this work, we proposed a tactics exploration framework (TEF) that applied grammar-based genetic programming (GP) to generating and evolving tactics in the engagementlevel simulation. Tactics are represented with modular behavior trees (BTs) for compatibility with the genetic operators. Experiments to explore submarine tactics have been conducted to observe and study the exploration process. The experimental results show that the TEF based on GP is efficient to explore tactics in the formalism of BTs.

show abstract

“…This process continued until a reliable script is reached. Both single player and coordinate team behaviors created by dynamic scripting were tested in [14,15]. Teng et al proposed FALCON [16] architecture based on self-organizing neural network to evolve strategies in 1-v-1 dogfight training simulation.…”

Section: Related Workmentioning

confidence: 99%

Adaptive CGFs Based on Grammatical Evolution

Yao

Huang

Wang

2015

Mathematical Problems in Engineering

View full text Add to dashboard Cite

Computer generated forces (CGFs) play blue or red units in military simulations for personnel training and weapon systems evaluation. Traditionally, CGFs are controlled through rule-based scripts, despite the doctrine-driven behavior of CGFs being rigid and predictable. Furthermore, CGFs are often tricked by trainees or fail to adapt to new situations (e.g., changes in battle field or update in weapon systems), and, in most cases, the subject matter experts (SMEs) review and redesign a large amount of CGF scripts for new scenarios or training tasks, which is both challenging and time-consuming. In an effort to overcome these limitations and move toward more true-to-life scenarios, a study using grammatical evolution (GE) to generate adaptive CGFs for air combat simulations has been conducted. Expert knowledge is encoded with modular behavior trees (BTs) for compatibility with the operators in genetic algorithm (GA). GE maps CGFs, represented with BTs to binary strings, and uses GA to evolve CGFs with performance feedback from the simulation. Beyond-visual-range air combat experiments between adaptive CGFs and nonadaptive baseline CGFs have been conducted to observe and study this evolutionary process. The experimental results show that the GE is an efficient framework to generate CGFs in BTs formalism and evolve CGFs via GA.

show abstract

Dynamic Scripting with Team Coordination in Air Combat Simulation

Cited by 17 publications

References 7 publications

Multi-Agent Multi-Objective Deep Reinforcement Learning for Efficient and Effective Pilot Training

Multi-Agent Multi-Objective Deep Reinforcement Learning for Efficient and Effective Pilot Training

Tactics Exploration Framework based on Genetic Programming

Adaptive CGFs Based on Grammatical Evolution

Contact Info

Product

Resources

About