Complex systems driven in whole or in part by artificial intelligence (AI) are fast becoming ubiquitous, across a broad range of applications. These systems must be tested and evaluated (T&E) to ensure they operate as specified, and not in undesirable ways. AI-based systems are inherently difficult to test because the underlying algorithms that drive them are used in extremely complex environments with overwhelmingly large combinations of inputs and variables. Nonetheless, engineers must perform T&E procedures for the systems to achieve a desired level of assurance, safety, and trustworthiness.
In this course, a team of experts in the field of AI and T&E will provide participants with an understanding of how cutting-edge T&E plans are developed for AI-driven autonomous systems. Participants also will engage in a variety of in-class exercises to practice the concepts of AI-based T&E for autonomous systems.
Participants in this artificial intelligence course will engage in a variety of interactive in-class exercises to get hands-on experience applying the latest in AI-based autonomous systems test and evaluation.
Describe the novel challenges introduced by AI, including machine learning (ML), in autonomous systems.
Describe the 6-D framework for creating an AI-enabled system, navigate AI technology solutions with an end-to-end engineering perspective, and explain how convolutional neural networks (CNNs) relate to the AI renaissance.
Summarize the basic math behind CNNs, compare CNNs to how the human brain works, and distinguish fact and fiction in CNN applications.
Identify the dimensions of complexity for autonomous systems, explain the role and requirements of simulations in autonomy development and testing, and identify the major architectural components of a modular autonomous system.
Describe a cyber-physical system; explain safety concerns related to AI-enabled cyber-physical systems; identify mitigations for addressing key classes of challenges and vulnerabilities.
Describe the challenges of confidence estimation for deep learning models and explain current approaches for dealing with problems of uncertainty and domain shift.
Identify metrics for performance, quality, reliability, safety for AI-enabled systems.
Test and Evaluation / Verification and Validation (TEVV) tools and methodologies for AI-enabled systems.
Identify challenges verification and validation on autonomous systems that perform control/planning tasks; describe approaches to formal verification of AI controllers; describe approaches safe design and fallback control architectures.
Describe the UL4600 standard and how it relates to MIL STD 882; lessons learned from R&D work in the industry.
Define the concept of explainable AI; explain how AI systems can be made more user-friendly; identify ways that failure modes can be made easier to understand.
List the essential elements of teamwork and describe how they might apply to AI-enabled and autonomous systems; explain how humans can interact safely and effectively with autonomous systems.
Describe strategies for human supervision of AI-enabled and autonomous systems; explain the overarching frameworks that apply to the governance of AI-enabled systems; Identify influence of human-systems interaction on assurance and trust.
Identify legal, policy, and ethical issues related to AI-enabled systems; list challenges of interpreting policy and ethical concepts for technology development.
Review case studies of testing and evaluating AI-enabled systems to identify lessons learned and potential pitfalls. Establish mock reviews of nominal AI-enabled systems to demonstrate application of learned tools to relevant challenges.
Identify key considerations for deploying autonomous systems in the real world.
A working knowledge of systems engineering practices and some experience with testing and evaluation of complex systems.