Planning with Self-Learning Models: A Practical Guide to Search, Control, and Decision Intelligence offers a clear and practical introduction to a new generation of reinforcement learning methods that combine learned world models with planning. The book explains how self-learning systems can use experience to build internal models of an environment, search over future possibilities, and make stronger decisions than purely reactive approaches. Readers are introduced to the central ideas behind model-based control, self-improving policy learning, and tree search, with an emphasis on intuition, mathematical foundations, and the design choices that make these systems effective in practice.
The book then moves into implementation, showing how to construct and train practical planning systems from the ground up. It covers representation learning, dynamics and prediction networks, uncertainty handling, optimization strategies, replay and data management, and the role of search in improving decision quality. Throughout, the text emphasizes stable training, scalable architectures, and robust evaluation, while also addressing common challenges such as partial observability, sparse rewards, computational cost, and generalization across changing environments. Step-by-step guidance, architectural patterns, and training recommendations make the material useful for both researchers and practitioners.
Beyond core methods, the book explores a wide range of applications in games, robotics, operations, autonomous systems, finance, and other domains where long-horizon planning and adaptive decision-making matter. It also examines emerging extensions such as stochastic modeling, hierarchical planning, meta-learning, hybrid control systems, and interpretable decision intelligence. By connecting theory, implementation, and real-world use cases, Planning with Self-Learning Models provides a practical roadmap for building intelligent systems that learn, search, and act effectively in complex environments.
