1,720,952 research outputs found
Using Causal Discovery to Design Agent-based Models
No full textDesigning agent-based models is a difficult task. Some guidelines exist to aid modelers in designing their models, but they generally do not include specific details on how the behavior of agents can be defined. This paper therefore proposes the AbCDe methodology, which uses causal discovery algorithms to specify agent behavior. The methodology combines important expert insights with causal graphs generated by causal discovery algorithms based on real-world data. This causal graph represents the causal structure among agent-related variables, which is then translated to behavioral properties in the agent-based model. To demonstrate the AbCDe methodology, it is applied to a case study in the airport security domain. In this case study, we explore a new concept of operations, using a service lane, to improve the efficiency of the security checkpoint. Results show that the models generated with the AbCDe methodology have a closer resemblance with the validation data than a model defined by experts alone.Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Air Transport & Operation
The Effect of Relationship Biases on AOCC Performance
No full textPassengers are frequently affected by airline disruptions, leading to a poorer than expected passenger experience. Airlines are affected by disruptions in the order of billions of dollars. Managing disruptions effectively is therefore paramount for an airline’s long-term commercial success. In spite of decision-support tools being introduced to facilitate Airline Operational Control Center (AOCC) decision-making, their adoption rate is low. For the foreseeable future, humans will unquestionably remain in the loop when it comes to AOCCdisruption management and human-factors will continue to come into play in AOCC decision-making. Toimprove AOCC decision-making, the effects of human factors on decision-making must be well understood. Bias is a human factor that affects decision-making and a relationship bias is a bias where previous negative experiences, between two individuals, will negatively affect future interactions they may have. A lack of trust, unwillingness to concede (in negotiations) or even a reluctance to interact, are a few examples on how a relationship bias may operationally manifest itself. AOCC decision-makers collaborate with one another to arrive at a integrated solution that mitigates an airline’s disruption. If a relationship biases exist within the AOCC, this negatively affects collaboration among AOCC decision-makers the development of solutions. The effect of the relationship bias on the solutions selected to mitigate an airline’s disruptions motivates the study of the effect of the relationship bias on AOCC performance. The fact that the relationship bias on AOCC decision-making has never been research, further motivates its study. We hope to address this research gap by evaluating the effects of the relationship bias on AOCC performance. More precisely, the research objective is to evaluate the effects of the relationship bias on AOCC performance, by modelling AOCC decision-making through a Naturalistic Decision-making framework using Klein’s Extended Recognition-Primed Decision model, and modelling AOCC social decision-making and interactionsusing Chow’s Co-Ladder model. The methodology involves formalizing AOCC goals through a framework [Popova and Sharpanskykh, 2008] which enables us to measure organizational performance. Furthermore, it involves formally integrating Bruce’s extension [Bruce, 2011a] of Klein’s extended Recognition-primed Decision (RPD) model with Chow’s social interaction model [Chow et al., 2000]. The model is finally simulated for a scenario where a scheduled flight suffers a mechanical disruption and the performance is evaluated based on a goal satisfaction and operational costs. There are three main contributions of this research. Firstly, the individual cognition model developed makes it possible to model AOCC decision-maker’s individual cognition within the complex and dynamic AOCC environment. The second contribution is the proposed integrated model, which make it possible to integrate agent individual cognition and agent social decision-making and interactions. The third contribution is the evaluation of various possible relationship biases, which makes it possible evaluate the effect of different relationship types on AOCC performance. The research conducted led to a few interesting findings. For example, only some relationship biases lead to a significant decreased in AOCC performance, whereas other relationships have a negligible effect. Another interesting finding was that if there is a relationship bais between two agent, they are both not equally affected. The relationship bias only affects the AOCC control agent who requires information from a counterpart in order to develop their partial solution.Aerospace Engineerin
Agent-Based Modelling for Security Risk Assessment
Full text linkSecurity Risk Assessment is commonly performed by using traditional methods based on linear probabilistic tools and informal expert judgements. These methods lack the capability to take the inherent dynamic and intelligent nature of attackers into account. To partially address the limitations, researchers applied game theory to study security risks. However, these methods still rely on traditional methods to determine essential model parameters, such as payoff values. To overcome the limitations of traditional methods, we propose an approach which combines agent-based modelling with Monte Carlo simulations. Agent-based models allow more realistic representation of essential aspects and processes of socio-technical systems at cognitive, social and organisational levels. Such models can be used to estimate risks and parameters related to them. An application of the approach is illustrated by a case study of an airport security checkpoint.Air Transport & Operation
A Machine Learning Approach to Flight Safety Event Prediction
No full textSafety occurrences in the aviation industry are nowadays commonly regarded as the outcome of a complex system. Due to this systemic view on safety airlines pursue to understand this complex, underlying system and aim to proactively act upon the occurrence of these events. The most prevalent implementation of flight safety event detection is however still threshold analysis, which has no such implications. On the other hand, Machine Learning methods have readily proven to be an efficient and valuable solutions in predicting the occurrence of anomalies in data, such as flight safety events. However, existing methods search for anomalies in datasets encompassing the anomaly, i.e. direct datasets. On the contrary, this study approached airline operations as a complex system which the outcome could be the occurrence of a flight safety event. Hence, the question was raised whether a set of indirect precursors could be significant in predicting flight safety events. That is why common airline processes were selected, in consultation with industry experts, and their indirect data considered. The aim of this study was to evaluate this concept by evaluating a set of precursors for a particular flight safety event (a case study). The Knowledge Discovery in Databases framework was the general guideline throughout this research, with a Relief and Neural Network algorithm as transformation and data mining step respectively. This study showed that the considered processes were significant in predicting the occurrence of a safety event, although the found precursors could not fully encompass the event under investigation. The classification performance of the methodology was characterised by a large number of false positives, which originated from the problem's class skewness. The Matthews Correlation Coefficient proved to be a well-balanced optimisation objective for such problems and overcame this drift. Locally, the weight optimisation showed a set of confidently classified false positives and negatives confined further improvement. These misclassifications were found to be the result of a lack of adequate information. Nevertheless, the considered information did display to be significant as the obtained Matthews Correlation Coefficient and recall underpinned, particularly in the light of the class imbalance and the anomalous nature of flight safety events.Aerospace Engineerin
Simulating an Scheduling Airport Security Checkpoints: Q-Learning-Based Allocation of Operators to Security Teams at an Airport Security Checkpoint
No full textIn this paper, we propose and analyze a q-learning-based approach for allocation of operators to security teams in order to improve operational efficiency of an airport security checkpoint. The research is composed of two parts. First, we develop an agent-based model capable of simulating an airport security checkpoint. Second, we introduce learning agents into the model, whose goal is to allocate operators to a security team during operations, to improve operational efficiency of the security checkpoint. We propose two learning activities of these agents. Activity 1 allocates operators to the recheck process, where operators are responsible for reexamining luggages that have been rejected and decide if they are safe or not. Activity 2 allocates operators to the CT process, where operators are responsible for examining CT images of luggages and decide if a luggage should be rechecked or not. We demonstrate that introducing a learning agent with either activity increases the throughput of the security checkpoint. Furthermore, activity 1 and activity 2 decrease the time spent in critical operations for the recheck process and CT process, respectively. The behavioural strategies learned by the agents were to add an operator when there is excess luggage waiting and remove an operator when there are excess operators available. Policy evolution between two different learning agents was compared by determining the similarity in their state transition networks per episode. The similarity was computed using the DeltaCon method and proved to be a promising technique for identifying differences in agent behaviour. This study appears to be the first to dynamically-schedule security operator shifts using a reinforcement learning approach. Insights gained from this study demonstrate that dynamically allocating operators to a security lane improves its operational efficiency, which opens the possibility of dynamic-scheduling security operators for entire terminals. Furthermore, it may aid airport managers in creating more resilient security checkpoints.Aerospace Engineerin
An Agent-Based Social Simulation Approach to Task Allocation in Aircraft Maintenance Teams
No full textTighter profit margins and rising aircraft complexity are currently driving the need for aircraft maintenance organizations to increase efficiency. Many organizations believe that digitization is key for improving operational performance. Digitization of the task allocation process at a large aircraft maintenance organization did unexpectedly not lead to increased efficiency. Anthropological research concluded that the implemented technologies did not accommodate for the social nature of teamwork. This research studies the relationship between the social aspects of teamwork and the performance of task allocation methods in aircraft maintenance. An Agent-Based Social Simulation model has been created and simulated for different types of task allocation methods as well as different types of teams. The presented model includes social influence relations between team members and decision-making based on trust in others' performance. The model has been simulated for a case study of an Airbus A310 main landing gear replacement. Independent teams provided the best performance using a mediated feedback automated negotiation method. The most beneficial results for compliant teams were obtained through task allocation by the team lead. Social teams presented significantly better results for voting than for other task allocation methods. The combination of socially oriented mechanics and task allocation by voting provided the most advantageous task execution performance for all simulations. It was shown that, in line with the wisdom of crowds theory, diversity in initial trust levels in combination with shared trust among mechanics over time increased collaborative task allocation performance.Aerospace Engineerin
An Agent-Based Safety Analysis of the Third Party Risk of UAS operations in the urban environment
No full textThere are many diverse concepts for autonomous drone operations. One of these concepts is the delivery of packages in urban areas. Transportation companies are already testing vehicles capable of performing such operations, and many studies indicate that this concept is both technically feasible and financially attractive. However, how much risk do civilians living in cities (also called third-party-risk, or TPR) face following these drone operations? Moreover, what measures can be employed to mitigate this risk? Regulators demand answers to these questions before allowing autonomous Unmanned Aircraft Systems (UAS) operations to take off. Recent research has focused on developing methods of calculating the TPR, or on creating models that accurately resemble drone operations in urban cities. The model proposed in this work bridges the gap between these categories using an agent-based safety risk analysis. In this analysis, we study the TPR of UAS package delivery operations in Delft, New York and Paris. This model leads to two main contributions. The first considers observations regarding the influence of the environment on the TPR. In particular, our model suggests that if the low-risk areas are clustered in a city, this leads to higher TPR. The second contribution is derived from the interaction of the risk computation with our model of the environment. Global- and local sensitivity analyses led to a few interesting observations. For example, our model suggests that it is more important to understand the vehicle's failure rate in the cruise-phase, than in the takeoff- and landing-phase. It is also suggested that it is more important to understand how buildings protect people from impact than to understand the effects of an impact directly on a human. Another finding is that modelling the impact speed with the terminal speed, as is common in literature, leads to a TPR that is 15% - 26% higher than when the impact speed is modelled based on the drone's dynamics.Aerospace Engineerin
Socio-Technical Agent-Based Modelling of Decision-Making in an Airline Operational Control Center: Application to an unexpected diversion use case at KLM CityHopper
No full textThis paper proposes a socio-technical agent-based model developed to increase the understanding of decision-making in an airline operational control center (AOCC). In this model human decision-makers, a Decision Support Tool, and technical systems are included. The selected case study encompasses theunexpected diversion of a flight from Amsterdam to London City. The decision-making inherent to the disruption management of this unexpected diversion is studied. In the model, in line with cognitive science literature, the agents are simulated to operate in the scrambled, opportunistic, tactical, and strategic con-trol modes, which differ in the situation awareness and use of technical systems. In these control modes respectively the plurality voting protocol, Borda voting protocol, Clarke Tax Algorithm, and Multi-Criteria Decision-Making (MCDM) are implemented as decision-making mechanisms. The agents in the strategiccontrol mode demonstrated the ability to make adaptive decisions. In the tactical control mode, the agents showed decision-making characterised by adaptive responses and limited anticipation. In the scrambled and opportunistic control modes, the decision-making was characterised by a lack of adaptation and was solely based on experience. The analysis of the decision-making showed that decision-making based on the airline’s cost model results in different decisions than decision-making by the human operations controllers. Due to this, the operations controllers are not eager to decide on the implementation of a proposed solution strategy by the Decision Support Tool. It showed that the decision-makers often have to make compromises to arrive at a collaborative decision. Furthermore, scenarios that are characterised by reserve unavailability do not require anticipation in the decision-making and can be handled in the opportunistic control mode, which is the lowest control mode that resulted in adequate decision-making. However, when anticipation is required the CDM and OrbiFly systems are essential resources. It is recommended that the MCDM decision-making mechanism can be used to improve the consistency of the operational decision-making and enables the AOCC to learn from previous occurrences. Hereby, the capability of the AOCC to deal with disruptionsthrough resistance instead of resilience can be extended. Overall, the analysis of the different decision-making mechanisms showed that human operations controllers are essential for adaptive decision-making in the AOCC.Aerospace Engineerin
An agent-based modelling study and analysis of an adaptive multi-UAV virus test delivery system
No full textIn February and March 2020, the COVID-19 disease spread rapidly across the world. It was a new disease for which many countries were not prepared. In the Netherlands there was a scarcity of virus tests, and lack of coordination on the allocation of these scarce tests. Besides, there was no proactive acquisition ofspread information, and spread predictions were not used. To address these issues, government institutions could use UAVs. The UAVs acquire virus spread information, by coordinated and proactive delivery of virus tests to people’s home. This information is important for detecting virus outbreaks early, and to respond with appropriate measures, such as social distancing measures. This research aims to develop and evaluate such a multi-UAV test delivery system, using the agent-based modelling and simulation paradigm. The system consists of two main components: (1) The system coordinates the UAVs’ region of focus, by using a Bayesian decision network. This Bayesian decision network uses the virus case predictions of a spatiotemporal generalised linear prediction model, and the observations of the UAV system to make coordination decisions. (2) The system uses neighbourhood search and exploration techniques to detect virus cases that otherwise would not be identified. The methodology has been applied to a case study of 5 municipalities in the province of Noord-Brabant in the Netherlands. These municipalities were strongly hit by the COVID-19 epidemic. The main conclusions are that increasing the number of UAVs, increases the virus detection and decreases average delivery time. Furthermore, the Bayesian decision network is effective in prioritising the allocation of test resources to regions with higher epidemic severity. It does so by allocating more time to severe regions. Additionally, neighbourhood search is an effective way to find unobserved cases. Moreover, exploration in combination with the fixed response threshold model was found to be more effective than neighbourhood search.Aerospace Engineerin
Agent-based Modelling and Simulation of Airport Terminal Operations under COVID-19-related Restrictions
No full textThe worldwide COVID-19 pandemic has had a tremendous impact on the aviation industry, with a reduction in passenger demand never seen before. To minimise the spread of the virus and to gain trust from the public in the airport operations' safety, airports implemented measures, e.g., physical distancing, entry/exit temperature screening and more. However, airports do not know what the impact of these measures will be on the operations' performance and the passengers' safety when passenger demand increases back. The goal of this research is twofold. Firstly, to analyse the impact of current (COVID-19) and future pandemic-related measures on airport terminal operations. Secondly, to identify plans that airport management agents can take to control passengers' flow in a safe, efficient, secure and resilient way. To model and simulate airport operations, an agent-based model was developed. The proposed model covers the main airport's handling processes and simulates local interactions, such as physical distancing between passengers. The obtained results show that COVID-19 measures can significantly affect the passenger throughput of the handling processes and the average time passengers are in contact with each other. For instance, a 20% increase in check-in time (due to additional COVID-19 related paperwork at the check-in desk) can decrease passenger throughput by 16% and increase the time that passengers are in contact by 23%.Aerospace Engineerin
- …
