This survey provides a comprehensive overview of imitation learning, a method where an agent learns to perform a task by mimicking a teacher. The authors, Nathan Gavenski, Felipe Meneguzzi, Michael Luck, and Odinaldo Rodrigues, from King’s College London, University of Aberdeen, University of Sussex, and King’s College London, respectively, address the challenges of comparing and evaluating different imitation learning techniques, environments, and metrics. They introduce novel taxonomies to categorize these aspects, reflecting on key issues and highlighting open challenges and future directions. Imitation learning is beneficial because it leverages samples from a proficient teacher, reducing the need for extensive trial-and-error or large supervised learning datasets. This approach is particularly useful when the desired behavior is difficult to specify through a reward function. The field has seen significant growth, leading to various techniques and applications, but the lack of standardization in evaluation methods poses challenges. The survey covers the background of imitation learning, including Markov Decision Processes (MDPs) and the role of demonstrations, experiences, and observations. It discusses different methods such as behavioral cloning, dynamics models (inverse and forward), and adversarial learning. Each method is evaluated based on its efficiency, sample efficiency, and generalization capabilities. Key contributions of the survey include: 1. **Taxonomies**: The authors propose new taxonomies for imitation learning methods, environments, and metrics, which are more detailed and complementary to existing classifications. 2. **Methodological Insights**: They provide a systematic review of 50 publications, highlighting the most common approaches and their applications. 3. **Challenges and Future Directions**: The authors discuss key insights, challenges, and potential future research directions, emphasizing the need for consistent evaluation processes and the use of new taxonomies. The survey aims to provide a solid foundation for understanding the diverse and dynamic field of imitation learning, facilitating the development of new approaches and more systematic comparisons of techniques.This survey provides a comprehensive overview of imitation learning, a method where an agent learns to perform a task by mimicking a teacher. The authors, Nathan Gavenski, Felipe Meneguzzi, Michael Luck, and Odinaldo Rodrigues, from King’s College London, University of Aberdeen, University of Sussex, and King’s College London, respectively, address the challenges of comparing and evaluating different imitation learning techniques, environments, and metrics. They introduce novel taxonomies to categorize these aspects, reflecting on key issues and highlighting open challenges and future directions. Imitation learning is beneficial because it leverages samples from a proficient teacher, reducing the need for extensive trial-and-error or large supervised learning datasets. This approach is particularly useful when the desired behavior is difficult to specify through a reward function. The field has seen significant growth, leading to various techniques and applications, but the lack of standardization in evaluation methods poses challenges. The survey covers the background of imitation learning, including Markov Decision Processes (MDPs) and the role of demonstrations, experiences, and observations. It discusses different methods such as behavioral cloning, dynamics models (inverse and forward), and adversarial learning. Each method is evaluated based on its efficiency, sample efficiency, and generalization capabilities. Key contributions of the survey include: 1. **Taxonomies**: The authors propose new taxonomies for imitation learning methods, environments, and metrics, which are more detailed and complementary to existing classifications. 2. **Methodological Insights**: They provide a systematic review of 50 publications, highlighting the most common approaches and their applications. 3. **Challenges and Future Directions**: The authors discuss key insights, challenges, and potential future research directions, emphasizing the need for consistent evaluation processes and the use of new taxonomies. The survey aims to provide a solid foundation for understanding the diverse and dynamic field of imitation learning, facilitating the development of new approaches and more systematic comparisons of techniques.