Mobile Augmented Reality Systems (MARS) e-learning has the potential to provide continuous, context-based, and autonomous instruction to human learners anytime, anyplace, and at any-pace. MARS e-learning enables mobility for the learner and hands free human computer interactivity. Advances to MARS based learning provides the advantage of a natural human-computer interface, flexible mobility, and context-aware instruction allowing learners to develop psychomotor skills while interacting with their natural environment with augmented perceptual cues. These perceptual cues combining multi-modal animation, graphics, text, video, and voice along with empirical instructional techniques can elegantly orchestrate a mobile instructional tool. The challenge, however, is building a MARS e-learning tool with capabilities for adapting to various learning environments while also considering the cultural, geographical, and other contexts about the learner. This paper discusses a novel system/software architecture, CAARS, for developing context-aware mobile augmented reality instructional systems.