In the development of actively controlled engineering systems, physical system (or plant) design decisions influence how associated control systems should be designed to maximize system utility, and control design decisions affect optimal plant design decisions. This property is known as bi-directional design coupling. Conventional sequential design strategies used in practice (i.e., plant design followed by control system design) do not fully-utilize opportunities for performance improvement available due to this coupling. A class of methods, referred to by some as control co-design (CCD) methods, do fully capitalize on the plant-control coupling to achieve system-optimal performance. CCD concepts and specialized CCD optimization strategies, such as earlier efforts in control-structure interaction, have been applied for some time. Recent advances in CCD have led to more comprehensive theoretical foundations, more generally applicable CCD methods, and strategies that manage unique challenges that arise from comprehensive treatment of physical system design, tools for balancing plant and control design complexity, realization of significant performance improvements, and emergence of non-obvious superior design solutions. Recently, CCD has been called by some decision makers an \`engineering game changer'.

A review of key milestones in CCD research will be presented, and current trends in CDD will be highlighted including the utilization of CCD for engineering applications of strategic societal (e.g., energy sustainability) and scientific (e.g., spacecraft attitude control) importance. Open challenges, such as incorporating comprehensive uncertainty treatment, implementable closed-loop control, and design-appropriate models will be explicated. The material presented here provides a basis and vision for the rapidly intensifying research activity at the interface of physical and control system design.

As an example, consider the design of a robotic manipulator. The physical-system design may comprise the links geometric properties while the control-system design may be embodied by the joint torque time trajectories for a specific task. Many authors have shown the benefit of a combined strategy rather than a sequential approach\~}. With a sequential approach, the plant is optimized initially, followed by the controller. The two primary CCD solution strategies that can produce system-optimal results are the simultaneous and nested (bilevel) optimization approaches. The simultaneous strategy optimizes both the plant and control variables in the same optimization formulation. With the nested strategy, an outer optimization loop optimizes the plant design, and an inner optimization loop identifies the optimal control for each candidate plant design desired by the outer loop. Recent work in the community has been expanding beyond these basic CDD strategies incorporating techniques from decomposition-based optimization, but guidelines and tools are still lacking. Comprehensive uncertainty treatment is an additional challenge because of the dynamic nature of CCD. When different problem elements are stochastic, appropriate treatment requires uncertainty distributions to be propagated with respect to a dynamic system. While specific techniques from fields such as robust optimization and control could be utilized, many have particular assumptions that need to be satisfied for their appropriate use, and such assumptions are frequently inappropriate in the CCD context. Under the systems perspective of CCD and bi-directional design coupling, it is desirable to explore different controller architectures when design freedom is available in the plant. Otherwise, optimal plant design solutions and system performance will be limited by predefined controller assumptions, and the desirable system behavior may not be observed. Therefore, many CDD studies have utilized optimal open-loop control to gain insights into the desired implementable, closed-loop controller. However, the process for realizing the desired implementable controllers is still quite ad hoc and informal. Hence, a \`\`gap'' certainly still exists in the current CCD methodologies and tools. Finally, the need for physical-design changes to be realizable is sometimes not a consideration in control design (because the plant is assumed to be fixed). Therefore, the development and effective usage of design-appropriate models in CCD is a critical aspect.

In this talk, we hope to highlight some successes, existing methods, and specific challenges of CCD research to this community.


Control Co-Design: Achieving New Functionality and Performance via Integrated Physical and Control System Design

Division / Sponsors: 
Design Engineering Division and Computers & Information In Engineering Division


Paper Proceedings: https://asmedigitalcollection.asme.org/IDETC-CIE/IDETC-CIE2020/volume/83907

IDETC2020 - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.