This paper presents an augmented reality system suited for maintenance training and support in industrial environments. We describe how augmented reality can help in reducing the effort of performing sequences of maintenance tasks in complex systems. The proposed approach is aimed to facilitate items localization in technical equipment by means of context-based instructions, virtual labels, 2D and 3D graphics, and animated virtual tools. These virtual contents are transformed to be visualized, co-registered to the real world, through a see-through Head-Mounted Display (HMD). To the aim of enhancing tracking reliability, we report some effective techniques to address drift errors in head tracking. The interaction paradigm is thought to reduce the keyboard usage to make the system less intrusive during user's activity. Deterministic finite automaton (DFA) equivalent representation is exploited to regulate maintenance assistance operations, providing a versatile mean to define either simple or complex procedure in a easy, verifiable and readable way. We also propose a collection of XML tags which enable the conversion of DFA in XML files, ensuring high extensibility and ease of understanding. First experimental results, targeted to augmentation of industrial racks, show a measurable improvement in performing maintenance operations with regard to the time needed to complete the task. © The Eurographics Association 2011.