Apex is a NASA Open Source Software architecture and development toolkit for creating intelligent, autonomous agents.

Apex is designed to lower the economic barriers to successfully deploying intelligent agents by reducing the time, expertise, and inventiveness required to build and maintain software agents. This means not only incorporating advanced artificial intelligence techniques, but also addressing numerous challenges to practical use of those capabilities that arise in complex, large-scale applications. This development philosophy, summarized by the project motto Usable Autonomy, has helped develop users both inside NASA and among researchers, teachers, and engineers.

The core component in Apex integrates artificial intelligence capabilities for monitoring events and for reactively selecting, scheduling, and controlling action. These capabilities are invoked by behavior representations specified by a user in a notation called Procedure Definition Language (PDL). Supporting these core mechanisms, the Apex Sherpa user interface combines components for building, debugging, analyzing and validating intelligent agent applications.

Apex has been used for a variety of purposes and in a variety of task domains including the following:

• Mission management and tactical flight control of an autonomous helicopter performing intelligent surveillance
• Simulating a Mars rover to help explore autonomy design parameters
• Simulating human air traffic controllers and pilots to help predict how people might respond to changes in equipment or procedures
• Predicting the precise duration and sequence of routine human behaviors (CPM-GOMS)
• At universities for teaching and student project work in courses on Cognitive Architectures and Human-System Engineering
• As the intelligent decision-making model underlying a dialogue management system
• As an artificial human subject in a collaborative decision-making experiment
• As a programmable human agent component of a more general airspace simulation "facility"
• Modeling and studying astronaut behavior and critical decision-making in space shuttle ascent phase

Experience with each application has helped drive Apex improvements aimed at making it easier to use for future applications. Many applications were developed by users outside the Apex team whose areas of expertise and experience with Apex vary greatly. Their experience and feedback has been particularly valuable in enabling iterative, user-centered refinement of Apex.

The guiding principle in developing Apex has been "usable autonomy" -- reducing the time, expertise, and inventiveness required to develop intelligent agent applications. Success at this requires going beyond the development of new and more powerful AI techniques to focus on, such as:

• Identifying and incorporating AI techniques of greatest value in the widest range of applications
• Making the behavior representation language expressive, intuitive, and readable
• Facilitating debugging and demonstration (Sherpa)
• Providing a flexible application architecture for multi-trial simulation, multiple agent control, interactive/exploratory simulations, and validation-driven iterative design.
• Providing good manuals, training materials, testbeds, and example applications
• Making it easy to acquire, install, try out, get help, and provide feedback on Apex
• Building infrastructure for native simulation and interoperation with other software
• Ensuring well-designed software: extensible, modular, robust, portable, and fast.