Automated Treatment Planning Systems (ATPS) have become one of the most prominent applications of artificial intelligence (AI)–driven transformation in radiation oncology. Advances in knowledge-based planning (KBP), deep learning (DL)–based dose prediction, and fully automated optimization engines over the past decade have enabled a transition from planner-dependent, manual workflows toward faster, more standardized, and highly reproducible radiotherapy planning processes. By learning from previously approved clinical plans, anatomical variations, and dose distributions, ATPS significantly reduce planning time, improve dosimetric consistency, and facilitate rapid plan adaptation, which is particularly critical for adaptive and online radiotherapy scenarios.

This work provides a comprehensive overview of the fundamental components of ATPS, including the principles and clinical implementation of KBP approaches and their commercially available solutions. Furthermore, DL-based automatic planning methods are discussed in detail, with a particular focus on voxel-wise dose prediction models employing architectures such as U-Net and DenseNet. These models leverage computed tomography (CT) images and structure segmentation masks to generate patient-specific, smooth, and clinically deliverable three-dimensional dose distributions. Beyond predicting final plan quality, DL-based dose estimates serve as strong priors for inverse planning, thereby improving optimization stability, reducing susceptibility to local minima, and enhancing overall planning efficiency.

The integration of automated contouring, beam geometry selection, DL-based dose prediction, optimization, and automated quality assurance (QA) into fully automated planning pipelines is also reviewed. In addition, modern optimization strategies for IMRT, VMAT, and proton and particle therapy are examined, highlighting the role of DL-based robust dose prediction and Bayesian optimization in managing range uncertainties and anatomical variability in proton therapy. Finally, the clinical performance, limitations, and future directions of ATPS are discussed, emphasizing the central role of AI-driven automated planning in advancing personalized, adaptive, and efficient radiotherapy.