Artificial intelligence meets embedded systems

Automation, optimization, and embedded system applications are strong competencies at PARC. These competencies are embodied in a unique discipline, "Embedded Reasoning." Embedded reasoning is an emerging multidisciplinary field driven by rapid advances in embedded computing power and the desire to control complex engineered systems reliably, efficiently, and safely. It tightly incorporates the strengths of AI reasoning — planning, scheduling, controlling, learning, and diagnosing — into physical systems. The end goal is to advance embedded system capabilities in solving complex tasks, acting on high-level goals, and adapting to dynamic and uncertain environments.

Recent success stories of autonomous systems operating in uncertain environments include NASA's Mars Exploration Rovers, the U.S. government's DARPA Grand Challenge, and unmanned aerial vehicles used in combat situations. In each case, the key to success is tight integration and collaboration between various engineering and computing disciplines. As advanced AI technologies find their way into autonomous systems, we will see a vast increase in the capabilities of such systems and the way they interact with the outside world.

Over the next decade, we envision:

• embedded reasoning methods will transform various application areas including robotics, transportation systems, defense, and industrial automation
• traditional interfaces of the fields of AI reasoning, control, and human factors will continue to blur as control systems and sophisticated AI algorithms are embedded in hardware
• machines will continue to gain attributes of intelligence including "self awareness," learning through experience, and context-sensitive behavior
• the proliferation of sensors in the physical world will lead to a proliferation of actuators that will manipulate the physical world in an autonomous fashion.

PARC's embedded reasoning work is based on competencies in automation and optimization, including:

• Model-based planning and scheduling
• Probabilistic modeling and reasoning
• Model-based diagnosis and state estimation
• Data mining and machine learning
• Prognostics and health management
• Model-predictive and optimal control
• Operations research
• Robotics

applications

Embedded reasoning is a thriving R&D area at PARC, with applications ranging from battlefield robots to aerospace vehicles, from cleantech to factory floors, and from printers to medical devices.

Fast Demand Response

Energy efficiency is a guiding principle for life in the 21st century. PARC efforts in cleantech include distributed automation and control technologies to reduce electrical power demand in a predictable and deterministic manner. PARC has developed patent-pending technologies for “fast demand response,” enabling sustainable demand reduction over wide areas at least an order of magnitude faster than existing commercial demand response approaches.

Prognostics and Health Management

Some of the pioneering work in model-based diagnosis was performed at PARC. PARC continues to push the boundaries of prognostics and health management technology through innovative work in pervasive diagnosis, data-driven prognostics, and model-based diagnosis applications for Xerox products. In addition, current PARC staff includes several leaders of the global prognostics and health management (PHM) as well as diagnostics communities.

Integrated Planning and Control

Over the last several years, PARC has developed a unique set of methods and tools for integration of model-based planning with control algorithms. The resulting PARC innovation, known as Plantrol, allows high-level planning functions to modify control behavior in real time through a lightweight negotiation protocol known as “Plantrol Contracts.” PARC’s patented Plantrol approach has found applications in diverse areas such as high-speed digital printers, manufacturing systems, and transportation systems.

Power System Optimization

PARC has developed unique optimization methods for managing power generation, storage, and utilization in embedded system applications. PARC competencies in power system optimization have applications in diverse areas such as vehicle design, smart grid management, and data centers.

Intelligent Robotics

Current PARC competencies in robotics include innovative control systems for autonomous quadrotor aircraft, distributed localization and mapping, and autonomous sensor networks.