Evgeny Kagan – författare

Plan optimal multi-robot search paths despite imperfect sensor information When multiple robots must locate targets in presence of false positive and false negative detection errors, path planning becomes extraordinarily complex. Multi-Agent Search under Uncertainty addresses this challenge directly. Written by researchers with combined expertise spanning defense systems, applied mathematics, and machine learning, this book delivers both theoretical foundations in search and screening theory and ready-to-use algorithms for practical implementation. The book covers cooperative search and navigation methods for autonomous mobile agents operating with incomplete or noisy information. Readers learn how Deep Q-Learning enables robots to develop complex behaviors through trial-and-error interactions rather than pre-programmed instructions. Applications span search and rescue operations, military surveillance, environmental monitoring, and security systems. An accompanying website provides Python code for simulation practice. Key topics include: Value-based Q-Learning methods where robots learn expected rewards for specific actions in given states under sensor uncertainty conditionsMulti-agent reinforcement learning approaches for swarm robotics where multiple robots learn cooperatively to accomplish collaborative search tasksDeep reinforcement learning using neural networks to process high-dimensional sensory inputs and execute complex search and tracking behaviorsAlgorithms for finding and tracking both stationary and moving targets while minimizing detection time despite false negative and positive readingsTheoretical contributions to search and screening theory alongside practical algorithms validated in autonomous robotic systems developmentDesigned for graduate students and researchers in robotics and reinforcement learning, this book bridges advanced theory with practical application. Professional developers building autonomous systems will find algorithms tested in real-world robotic development.

Plan optimal multi-robot search paths despite imperfect sensor information When multiple robots must locate targets in presence of false positive and false negative detection errors, path planning becomes extraordinarily complex. Multi-Agent Search under Uncertainty addresses this challenge directly. Written by researchers with combined expertise spanning defense systems, applied mathematics, and machine learning, this book delivers both theoretical foundations in search and screening theory and ready-to-use algorithms for practical implementation. The book covers cooperative search and navigation methods for autonomous mobile agents operating with incomplete or noisy information. Readers learn how Deep Q-Learning enables robots to develop complex behaviors through trial-and-error interactions rather than pre-programmed instructions. Applications span search and rescue operations, military surveillance, environmental monitoring, and security systems. An accompanying website provides Python code for simulation practice. Key topics include: Value-based Q-Learning methods where robots learn expected rewards for specific actions in given states under sensor uncertainty conditions Multi-agent reinforcement learning approaches for swarm robotics where multiple robots learn cooperatively to accomplish collaborative search tasks Deep reinforcement learning using neural networks to process high-dimensional sensory inputs and execute complex search and tracking behaviors Algorithms for finding and tracking both stationary and moving targets while minimizing detection time despite false negative and positive readings Theoretical contributions to search and screening theory alongside practical algorithms validated in autonomous robotic systems development Designed for graduate students and researchers in robotics and reinforcement learning, this book bridges advanced theory with practical application. Professional developers building autonomous systems will find algorithms tested in real-world robotic development.

Multi-valued and fuzzy logics provide mathematical and computational tools for handling imperfect information and decision-making with rational collective reasoning and irrational individual judgements. The suggested implementation of multi-valued logics is based on the uninorm and absorbing norm with generating functions defined by probability distributions. Natural extensions of these logics result in non-commutative and non-distributive logics. In addition to Boolean truth values, these logics handle subjective truth and false values and model irrational decisions. Dynamics of decision-making are specified by the subjective Markov process and learning – by neural network with extended Tsetlin neurons. Application of the suggested methods is illustrated by modelling of irrational economic decisions and biased reasoning in the wisdom-of-the-crowd method, and by control of mobile robots and navigation of their groups.Topics and features:Bridges the gap between fuzzy and probability methodsIncludes examples in the field of machine-learning and robots’ controlDefines formal models of subjective judgements and decision-makingPresents practical techniques for solving non-probabilistic decision-making problemsInitiates further research in non-commutative and non-distributive logicsThe book forms a basis for theoretical studies and practice of decision-making under uncertainty and will be useful for computer scientists and mathematicians interested in multi-valued and fuzzy logic, as well as for engineers working in the field of data mining and data analysis.

Multi-valued and fuzzy logics provide mathematical and computational tools for handling imperfect information and decision-making with rational collective reasoning and irrational individual judgements. 

The suggested implementation of multi-valued logics is based on the uninorm and absorbing norm with generating functions defined by probability distributions. Natural extensions of these logics result in non-commutative and non-distributive logics. In addition to Boolean truth values, these logics handle subjective truth and false values and model irrational decisions. Dynamics of decision-making are specified by the subjective Markov process and learning – by neural network with extended Tsetlin neurons. Application of the suggested methods is illustrated by modelling of irrational economic decisions and biased reasoning in the wisdom-of-the-crowd method, and by control of mobile robots and navigation of their groups.

Topics and features:

Bridges the gap between fuzzy and probability methodsIncludes examples in the field of machine-learning and robots’ controlDefines formal models of subjective judgements and decision-makingPresents practical techniques for solving non-probabilistic decision-making problemsInitiates further research in non-commutative and non-distributive logics

The book forms a basis for theoretical studies and practice of decision-making under uncertainty and will be useful for computer scientists and mathematicians interested in multi-valued and fuzzy logic, as well as for engineers working in the field of data mining and data analysis.