Complex decision-making processes, especially those associated with NP-hard optimization problems, become even more challenging when faced with uncertainty and dynamic conditions. Traditional approaches like simulation and heuristics provide effective but limited solutions, often unable to adapt to changing environments. In this paper, we introduce sim-learnheuristics, a methodological tool combining simulation, metaheuristics, and machine learning. Sim-learnheuristics employ simulation models to model uncertainty, metaheuristic algorithms to efficiently explore the solution space, and machine learning methods to guide and adapt the heuristic process based on evolving conditions. We describe the methodology, provide pseudo-code, and discuss the main parts of an implementation that illustrates the adaptability of this approach. A review of related work is included, showing the potential of sim-learnheuristics to improve decision making in real-life scenarios under dynamic and stochastic conditions.
The use of drones for routing and monitoring tasks has grown significantly, with applications such as traffic surveillance and road inspections gaining prominence . These real-world scenarios often involve unpredictable factors like fluctuating service times, which add complexity to traditional routing problems. This paper introduces a simulation-optimisation framework for routing drones under realistic conditions . To efficiently solve this problem, we propose a simheuristic approach that integrates a biased-randomised iterated local search metaheuristic with Monte Carlo simulation. Our computational experiments validate the efficiency, robustness, and speed of the proposed method, providing high-quality solutions to routing challenges in uncertain environments.
Network simulation tools provide advanced simulation solutions. These tools reduce costs and simplify the testing of scenarios to mitigate risks before deployment. However, many of these tools still possess shortcomings that hinder their usability and widespread adoption. Notably, many tools cannot accommodate dynamic scenarios, thus limiting the scale and complexity of simulations they can support. Our study explores professionals’ experiences with network simulation tools, with a focus on CORE/EMANE, aiming to understand their shortcomings and future potential. While existing research highlights technical drawbacks, our mixed-methods approach, uses a questionnaire to gather perspectives on network simulators and a design workshop to explore both technical and usability barriers associated with CORE/EMANE from an industry perspective. Our research highlights critical technical and usability barriers and challenges as perceived by industry professionals who frequently use these tools. Our work improves the understanding of the technical and usability requirements of network simulation tools. More importantly, we provide features that are grounded in participants’ feedback, aiming to streamline user interfaces and align tools capabilities with the evolving network simulation landscape for improved project outcomes. These insights are key for informing future development, emphasising the need for tools that address technical requirements while ensuring user-friendliness.
The credibility of simulation models for freight transportation builds on solid verification and validation processes. They are essential to ensure that simulation models accurately represent the systems they intend to mimic and produce reliable outcomes. Consequently, this paper reviews the documentation of verification and validation techniques in more than one hundred discrete event and agent-based simulation publications from the past decade to identify relevant methods and their frequency of application. Future simulation research on verification and validation approaches and helpful guidelines are needed to ensure rigorous implementation in logistical modelling.
In the realm of autonomous robotics, particularly in disaster response scenarios, the ability to efficiently visualize swarm behaviors and decision-making processes is crucial for development, testing, and real-time operation. This paper delves into the autonomy core system developed for combined virtual and real robot swarms, emphasizing the modular architecture that allows for seamless integration of different visualization methods without necessitating significant code alterations. We explore the autonomy system’s design, its benefits for flexibility in visualization, and potential applications across various domains of robotics.
