Knowledge-based augmented reality (KARMA) is a system that enhances the real world with virtual information to assist users in performing complex 3D tasks. Unlike virtual reality, which replaces the real world with synthesized content, augmented reality enriches the real world by providing annotations, descriptions, and instructions. KARMA uses a head-mounted display and a knowledge-based graphics system to create personalized, private augmented realities that are not limited to traditional displays or interaction devices. The system is based on the IBIS (Intent-Based Illustration System), a rule-based system that generates illustrations to satisfy communicative goals. IBIS distinguishes between design and style, with design focusing on the high-level structure of the illustration and style on the specific visual effects. IBIS uses methods and evaluators to design and assess illustrations, allowing it to handle dynamic constraints and goals. KARMA is tested with a simple end-user maintenance application for a laser printer, where it helps users perform tasks like refilling the paper tray and replacing the toner cartridge. The system uses a see-through head-mounted display and a 3D position and orientation-tracking system to provide real-time feedback and guidance. KARMA's architecture includes a display server, object servers, and a head server that work together to maintain the integrity of the object and head motion information. The system processes data from motion trackers and uses a knowledge-based approach to generate dynamic, interactive augmented realities that respond to user interactions and changes in the real world. The system's design allows for the creation of rich, interactive augmented realities that support complex tasks by highlighting objects, providing labels, and guiding users through actions. KARMA's graphics component is designed to handle visibility constraints and ensure that objects remain visible and identifiable in the augmented reality environment. Future work includes improving the accuracy of the system, enhancing the user interface with speech and audio, and exploring the use of more advanced display technologies. The research also focuses on developing formal models of user performance and improving the visibility determination in augmented reality systems.Knowledge-based augmented reality (KARMA) is a system that enhances the real world with virtual information to assist users in performing complex 3D tasks. Unlike virtual reality, which replaces the real world with synthesized content, augmented reality enriches the real world by providing annotations, descriptions, and instructions. KARMA uses a head-mounted display and a knowledge-based graphics system to create personalized, private augmented realities that are not limited to traditional displays or interaction devices. The system is based on the IBIS (Intent-Based Illustration System), a rule-based system that generates illustrations to satisfy communicative goals. IBIS distinguishes between design and style, with design focusing on the high-level structure of the illustration and style on the specific visual effects. IBIS uses methods and evaluators to design and assess illustrations, allowing it to handle dynamic constraints and goals. KARMA is tested with a simple end-user maintenance application for a laser printer, where it helps users perform tasks like refilling the paper tray and replacing the toner cartridge. The system uses a see-through head-mounted display and a 3D position and orientation-tracking system to provide real-time feedback and guidance. KARMA's architecture includes a display server, object servers, and a head server that work together to maintain the integrity of the object and head motion information. The system processes data from motion trackers and uses a knowledge-based approach to generate dynamic, interactive augmented realities that respond to user interactions and changes in the real world. The system's design allows for the creation of rich, interactive augmented realities that support complex tasks by highlighting objects, providing labels, and guiding users through actions. KARMA's graphics component is designed to handle visibility constraints and ensure that objects remain visible and identifiable in the augmented reality environment. Future work includes improving the accuracy of the system, enhancing the user interface with speech and audio, and exploring the use of more advanced display technologies. The research also focuses on developing formal models of user performance and improving the visibility determination in augmented reality systems.