Research Programs

Can the system establish what is present, where it is, and which observations deserve confidence? This program addresses identity, localization, sensor reliability, and physical state estimation. Its scope ends at the validity and meaning of the measured state; diagnostic conclusions, control decisions, and adversarial attacks connect to the relevant neighboring programs.

What does the available evidence suggest about equipment condition or product quality? This program investigates deterioration, defects, diagnostic explanations, and likely root causes. It owns health and quality inference. Measurement validity belongs with Identification and Sensing, while process intervention and mobile inspection platforms connect to control and robotics research.

Which response serves an operational objective while respecting limits and permissions? This program covers industrial AI agents, optimization, planning, and control execution. It examines command preconditions and physical responses, with building and energy control among the research areas. Simulation infrastructure, broader assurance methods, and authorization security remain with their respective programs.

What can a digital experiment establish before a physical trial? This program addresses model fidelity, simulation environments, and evidence of transfer to equipment. It supports repeatable evaluation through calibrated models and hardware interfaces. The control and robot policies evaluated within those environments remain the responsibility of the programs developing them.

How should industrial robots, drones, and vehicles perceive, plan, manipulate, and coordinate? This program owns platform autonomy and fleet behavior, including intelligence distributed across connected devices. Supporting localization, simulation environments, assurance methods, and fleet security connect to the specialist programs responsible for those questions, with interfaces defined for each proposed study.

How can evidence show that behavior remains within stated limits when people and communication links are involved? This program addresses runtime assurance, outcome evaluation, operator interaction, and connectivity dependent behavior. Control policies belong with AI Decisions and Control; adversarial threats and security controls belong with Security and Trusted Industrial AI.

How could an adversary corrupt observations, misuse an AI agent, or issue an unauthorized physical command? This program owns adversarial threat models and security controls across sensing, decisions, and action. It also examines privacy, resilience, and recovery. Routine sensor faults, general assurance, and ordinary model lifecycle operations remain separate research responsibilities.

What is needed to reuse and maintain a research capability across industrial settings? This program addresses semantic integration, learning workflows, monitoring, release processes, and evidence of deployment value. Domain programs retain ownership of their scientific models and applications; malicious compromise is addressed through security research rather than routine lifecycle management.