Collaborative multiagent reinforcement learning schemes for air traffic management

Spatharis, Chistos; Blekas, Konstantinos; Bastas, Alevizos; Kravaris, Theocharis; Vouros, George A.

doi:10.1109/iisa.2019.8900719

Cited by 2 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is recommended to use separate training/test scenarios (with varying levels of stochasticity) while ensuring some similarities [66]. This study proposes a simplistic model with only two agent types and a static environment, with relatively short training time than other large-scale models [23,26,27]. Therefore, overfitting is less likely to occur and could be identifiable when individual behaviours are quantitatively and qualitatively interpreted.…”

Section: Proximal Policy Optimisation (Ppo)mentioning

confidence: 99%

“…The visually observed behaviours from the model are utilised, as these are the best indicators of intelligent adaptive behaviours in the model, as has been the case in past literature [30,73,50,23,26,72].…”

Section: Individual Behaviour Analysismentioning

confidence: 99%

“…Generally, RL applications in modelling can be split into two categories. In category (1), we have studies that adopt RL in their simulation to identify the optimal solution to a particular problem [23,26,27,49,50]. In category (2), we have studies that use RL as a decision-making framework to investigate a known phenomenon and observe new insights from this phenomenon, that may have not previously been identified due to constrained decision-making frameworks [24,45,52,53,54,55].…”

Section: Simulating Behaviour Using Reinforcement Learningmentioning

confidence: 99%

“…[24] further developed Schelling's model, a well-known ABM which explores segregation dynamics [25], utilising an RL decision-making framework to capture novel insights, such as evidencing that segregated areas are more likely to host older people than diverse areas, which attract younger ones. [26] harnessed the group-learning ability of RL (multi-agent systems where agents learn collectively as a group) to develop collaborative management dynamics of air traffic control. Similarly, [27] applied RL to develop policies for service management of cloud computing infrastructure.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Learning Complex Spatial Behaviours in ABM: An Experimental Observational Study

Olmez¹,

Birks²,

Heppenstall³

2022

Preprint

View full text Add to dashboard Cite

Capturing and simulating intelligent adaptive behaviours within spatially explicit individual-based models remains an ongoing challenge for researchers. While an ever-increasing abundance of realworld behavioural data are collected, few approaches exist that can quantify and formalise key individual behaviours and how they change over space and time. Consequently, commonly used agent decision-making frameworks, such as event-condition-action rules, are often required to focus only on a narrow range of behaviours. We argue that these behavioural frameworks often do not reflect real-world scenarios and fail to capture how behaviours can develop in response to stimuli. There has been an increased interest in Machine Learning methods and their potential to simulate intelligent adaptive behaviours in recent years. One method that is beginning to gain traction in this area is Reinforcement Learning (RL). This paper explores how RL can be applied to create emergent agent behaviours using a simple predator-prey Agent-Based Model (ABM). Running a series of simulations, we demonstrate that agents trained using the novel Proximal Policy Optimisation (PPO) algorithm behave in ways that exhibit properties of real-world intelligent adaptive behaviours, such as hiding, evading and foraging.

show abstract

Section: Proximal Policy Optimisation (Ppo)mentioning

confidence: 99%