Chapter Four - Situation-based genetic network programming to solve agent control problems

Created by W.Langdon from gp-bibliography.bib Revision:1.7686

@InCollection{ROSHANZAMIR:2024:ANCOT,

• author = "Mohamad Roshanzamir and Mahdi Roshanzamir",
• title = "Chapter Four - Situation-based genetic network programming to solve agent control problems",
• editor = "Anupam Biswas and Alberto Paolo Tonda and Ripon Patgiri and Krishn Kumar Mishra",
• series = "Advances in Computers",
• publisher = "Elsevier",
• volume = "135",
• pages = "77--97",
• year = "2024",
• booktitle = "Applications of Nature-Inspired Computing and Optimization Techniques",
• ISSN = "0065-2458",
• DOI = "doi:10.1016/bs.adcom.2023.11.003",
• URL = "https://www.sciencedirect.com/science/article/pii/S0065245823000864",
• keywords = "genetic algorithms, genetic programming, Evolutionary algorithms, Genetic network programming, Directed graph, Agent control problems",
• abstract = "Evolutionary algorithms are often used to generate the best solution based on a population of initial solutions and through successive generations. These algorithms have different types. One of them is Genetic Network Programming (GNP). This algorithm is one of the new evolutionary algorithms in which the structure of each individual is defined as a directed graph. This directed graph can be considered as a flowchart or strategy that can be used by the agent(s) to make decisions in the environment. So, this algorithm can be used for the automatic generation of solutions for agent control problems. Using GNP, researchers try to find the best individual (strategy) for an agent. However, if there is more than one agent in the environment, it is not easy to find a strategy that can optimally achieve the goal if all agents behave according to it. In this research, instead of looking for an optimal strategy for all agents, separate strategies are created for each agent based on its situation. This way, the goal can be achieved more easily and quickly because finding a strategy that can guide all the agents in final goal achievement is more difficult than finding different strategies that are created based on the situation of each agent. Generating different strategies gives more flexibility to the GNP algorithm for finding better solutions. For this purpose, the situation-based GNP (SB-GNP) algorithm has been proposed, which generates a strategy for each agent based on the situation of that agent. The results of applying the proposed method on Tile-World as a benchmark show that this method can improve the performance of traditional GNP. An important advantage of this method is that it can be added to all versions of the GNP without additional overhead",

}

Genetic Programming entries for Mohamad Roshanzamir Mahdi Roshanzamir

Citations