A CPS Approach to Robot Design
Abstract:
Analytically predicting the behavior of physical systems is generally not possible. For example, the three dimensional nature of physical systems makes it provably impossible to express closed-‐form analytical solutions even simple systems. This limitation makes experimentation the primary modality for designing new cyber-‐physical systems. Since the physical prototyping cyber-‐physical systems is typically very costly and hard to conduct, ``virtual experiments'' in the form of modeling and simulation can dramatically accelerate innovation in CPS development. Unfortunately, major technical challenges often impede the effective use of modeling and simulation in CPS design. This project is focused on developing better foundations and tools for modeling and simulating physical systems in robotics, which we believe is representative of many other CPS domains. To this end, the project consists of four activities: 1) Compiling and analyzing a benchmark suite for modeling and simulating robots, 2) Developing a meta-‐theory and a test bed for robot modeling and simulation, 3) Validating the research results of the project using two state-‐of-‐the-‐art robot platforms 4) Developing course materials incorporating the project's research results and test bed. During the two years, the project team:
(a) Developed a better understanding of the limitations of existing tools such as MATLAB, LabVIEW, and Simulink;
(b) Demonstrated that a small hybrid modeling language (Core Acumen) can be
highly expressive in capturing models for a variety of CPS systems;
(c) Devised multiple formalizations of a simulation semantics that computes a validated enclosure for hybrid systems, including ones that exhibit certain kinds of Zeno behavior;
(d) Developed a better understanding of the process involved in developing one
novel robot system;
(e) Demonstrated that exact real-‐arithmetic and other forms of self-‐validating real arithmetic (such as affine arithmetic and interval arithmetic) have the potential for serving as the foundations for the new more precise approach to modeling and simulation in CPS.
(f) Developed a new CPS course using Core Acumen that has been initially taught in the Embedded and Intelligent Systems (EIS) Masters program at Halmstad University.
More information about the project and it's software test bed (Acumen)
can be found at the web-‐page www.effective-‐modeling.org.