File preview
An Interdisciplinary Controls Curriculum for Cyber-Physical Systems Education
Patrick J. Martin
Electrical and Computer Engineering Program York College of Pennsylvania York, Pennsylvania 17403 Email: pjmartin@ycp.edu
Abstract—This paper presents an undergraduate controls curriculum that couples cyber-physical systems concepts with modern control design. The curriculum introduces cyber-physical systems topics to electrical and mechanical engineers that generally have less background in computing topics, such as concurrency and networking. This curriculum design increases undergraduate students’ awareness of cyber-physical systems.
• • • •
an ability to apply mathematical models of physical systems, cyber systems, and their composition an ability to design and conduct simulations and tests of a cyber-physical system and to analyze the results an ability to identify, formulate, and solve engineering problems that have both cyber and physical aspects a recognition for the need for, and ability to engage in life-long learning.
I.
I NTRODUCTION
Cyber-physical systems (CPS) require the application of many topics within modern controls research. Future controls engineers will be required to bring in additional concepts from communications and computing to solve these complex control problems [1]. There is a rising need to teach engineering students about the complexity introduced into modern control systems that utilize embedded computers and networks. This paper presents a new controls curriculum for undergraduates that incorporates several important topics from CPS with modern control design. At the undergraduate level, control systems is often stereotyped as a math-heavy, abstract subject. Consequently, [2] called for more integration of laboratory work into the controls curriculum. Since this work, several authors have investigated making controls more hands-on with laboratory work (e.g. [3], [4]). These lab activities give students valuable experience in the implementation of controllers on real systems. Many modern controllers are implemented in embedded computers, which requires students to know more about computing constraints, such as computational time and network delay. The work of [5] and [6] experimented with giving undergraduates experience in these topics, which are relevant topics in CPS education. These approaches create good lab experiences for students; however they should be accompanied by a course that blends the cyber and physical domains more readily. One example of a CPS focused curriculum is the work in [7]. This text provides a structure for introducing control principles to students in the computational sciences. However, there is still a need for the converse perspective: introducing computing and communications principles to students in the physical sciences and engineering. This paper presents an interdisciplinary, undergraduate controls curriculum that targets some of the proposed CPS learning outcomes:
To approach these learning outcomes, the curriculum blends modern control design with concepts from hybrid systems, networking, and concurrency. Since the students come from different engineering backgrounds, this curriculum must balance between physical and cyber systems topics to minimize student overload. At York College, electrical and mechanical engineers can take a two course sequence in control theory: Automatic Control and Applied Control. Automatic Control is a standard first semester course in control theory that covers classical topics such as frequency-domain modeling and analysis techniques. Near the end of the course students learn about PID control and apply use it to set the position of a DC motor using analog electronics. The sequel course, Applied Control, in the past was taught as a digital control course. However, mechanical engineers do not have the same background in digital signal analysis or programming as the electrical engineers. The overhead of teaching digital signal analysis and low-level programming distracted students from focusing on the control design concepts and tools. More recently, model-based design of controllers has become more common in the workplace. Products from Mathworks and National Instruments facilitate control design by providing tool chains that automatically generate low level code. Applied Control was re-designed for Spring 2013 to use the curriculum presented in this paper. All students in the course are familiar with feedback control; however, their knowledge of cyber systems varies. Thus, the course aims to introduce the cyber topics through project work and lectures. Students that take this course will be able to operate as control engineers and, furthermore, have a better perspective on the interplay between cyber and physical domains.
II.
C URRICULUM S TRUCTURE
B. Computing Models and Implementation The coupling of sensors, actuators, embedded computers, and networks presents many challenges for future control engineers. Students that do not have a strong background in computer science or engineering should study the cyber domain so that they can design high level control architectures. These architectures will need to take into account the constraints and abilities of cyber systems. For example, students interested in pursuing robotics need to be aware of discrete dynamics caused by computers switching among behaviors within an embedded operating system. At this point in the curriculum, students are introduced to finite state machines and hybrid systems. These topics develop their ability to apply mathematical models for cyber and physical systems. Interactive simulations of hybrid systems let students see the interesting effect hybrid control strategies have on physical dynamics. In model-based control design software, functions are represented as individual actors that process inputs and generate outputs. One advantage of this software design approach, is the ability to run concurrent control tasks with minimal thread locking issues. Students need to examine the principles of concurrency in actor and data-flow models so that their control implementations work across asynchronous computation elements. Exposing students to this theory is important for CPS education as more embedded systems are shipping with multicore processors. CPS engineers will be required to design systems that can balance power consumption, performance, and correctness. Thus, students should be aware of competing control loops in CPS applications. Consider the case where elements within a sensor network may need to conserve individual energy consumption; however, they may also be required to stream video data at a maximum rate. Recognizing physical and cyber system limitations in systems such as the example is an important skill when deploying CPS. This educational goal is best accomplished through an open-ended project that combines the foundational tools from control and computing. C. Project-Based Learning for CPS Most topics within this curriculum can be approached with lectures, in-class activities, and laboratories. However, CPS is a rapidly advancing field and students must synthesize new theories and technologies as they are developed. A projectbased learning activity (e.g. [8], [9]) complements this curriculum by providing a capstone experience for students to apply theory to solve a non-trivial problem. Project-based learning facilitates interdisciplinary team work, which is an important skill for students interested in developing CPS. Additionally, it provides motivation for students to reach outside of their academic comfort zone, which encourages their lifelong learning abilities. As an example of project-based learning for CPS, students in Applied Control at York College are assigned a project that utilizes the theoretical tools presented in II-A and II-B. The project involves the design and implementation of a multiagent solar farm. Each team of students is given a set of NI
An interdisciplinary controls curriculum for undergraduates needs to strike a balance between the students’ pre-requisite knowledge and the skills needed by the cyber-physical systems community. This particular curriculum builds student confidence in fundamental control system analysis and design topics before layering in the necessary CPS topics. Students apply their control design skills within simulation and hardware labs, which helps build intuition about state feedback and observers. They transition to CPS topics by studying finite state machines and hybrid systems, since they facilitate the modeling of cyber systems. Additionally, they learn about the basics of concurrency and networking and how they are used in modern control systems. A project-based learning activity is included so that students synthesize many of the topics within the curriculum into an application.
A. Foundations in Control Before students learn how to develop controllers for cyberphysical systems, they require a solid foundation in the core principles of modern control design. Students develop their intuition about physical systems by analyzing their models for stability. Additionally, students explore stabilization and regulation by applying classical PID control. PID control is a natural point of entry for the discussion of cyber systems topics, such as communication delay. To illustrate this topic, students explore the effects that delay has on DC motor performance when it is controlled over a network. They implement a PID controller on a laptop that issues control commands over a network interface to a real-time system that simulates a motor model. CPS are embedded within the physical world and are naturally multi-input, multi-output systems. Consequently, teaching state space analysis and design gives students a foundation for thinking about more complex system interaction. Students can apply controllability and observability to understand the structure of the system they are trying to control, and possibly re-design the system itself. These topics allow students to think about the actuation and sensing requirements of CPS applications. Students also learn about state feedback and its complementary topic of state observation for system regulation. Typically, these topics are reserved for graduate courses; however, introducing these tools in an undergraduate curriculum assists in developing intuition before invoking rigorous mathematical proof. Once students are familiar with these modern control techniques, they learn about optimal control through the Linear quadratic regulator (LQR) problem. Students experience these modern control design tools through LabView1 and Matlab2 simulation labs. Model-based design is emphasized so that students can focus on designing controllers. By following this foundational sequence, students become proficient at applying fundamental state-space analysis and design before adding on the computation and communication constraints in cyber systems.
1 http://www.ni.com/labview/ 2 http://www.mathworks.com/products/matlab/
Compact RIO computers3, one of which is shown in Figure 1. Each of these Compact RIO systems is interfaced to a Quanser SRV-02 servo motor4 , as shown in Figure 2.
advanced computing topics. However, undergraduate CPS education does not need to make every engineer an expert in these topics. Instead, it should make various types of undergraduate engineers aware of the field of CPS and the current challenges facing the research and industrial communities. Since this curriculum was deployed in the Spring 2013 semester, there is no assessment data to confirm if the curriculum structure satisfies the proposed outcomes. The author plans to take surveys at the conclusion of the course to measure student achievement with these learning outcomes. Some casual conversations with students indicate that their interest in the CPS-related topics is holding, despite the material being outside of their engineering specialty.
Fig. 1. A National Instruments Compact RIO computer with Quanser Q1USB module.
Additionally, the author will investigate the feasibility of moving some topics earlier in the undergraduate engineering program. A re-designed two-course sequence may be better suited to develop an undergraduate engineer’s appreciation for feedback control, computation, and communication. III. C ONCLUSION
Fig. 2.
A Quanser SRV-02 motor and its associated power amplifier.
The students must design a multi-agent control system where each agent actuates a solar panel (shown in Figure 3). The goal of the multi-agent system is to produce maximum total power generated by the solar panels while minimizing the control effort at each solar panel actuator. The agents operate independently and communicate over an ethernet network using TCP/IP. This project has multiple domain challenges for the students, from mechanical design and system modeling to algorithm selection and networked control. Furthermore, it highlights competing control design goals: energy efficiency and performance. To design their control system, the students need to perform research on current topics in networked and multi-agent systems as well as power systems.
This paper presented a controls curriculum that incorporates topics in cyber-physical systems. In addition to learning modern control design, students learn about concepts in computing that are outside of their major. Furthermore, they experience realistic physical and cyber constraints through a projectbased learning activity. This curriculum lays a foundation for undergraduate students to contribute to the CPS community in their future careers and research. ACKNOWLEDGMENT The author would like to thank the students of York College’s Applied Control course for participating in this experimental curriculum. R EFERENCES
[1] Ragunathan Raj Rajkumar, et al. “Cyber-physical systems: the next computing revolution” Proceedings of the 47th Design Automation Conference, ACM, 2010. Bernstein, Dennis S. “Enhancing undergraduate control education.” Control Systems Magazine, pp. 40-43, 1999. Arz´ n, K-E., Anders Blomdell, and Bj¨ rn Wittenmark. “Laboratories e o and real-time computing: integrating experiments into control courses.” Control Systems Magazine, vol. 25, no. 1, pp. 30-34, 2005. Bonnie S. Heck, N. Scott Clements, and Aldo A. Ferri. “A LEGO experiment for embedded control system design.” Control Systems Magazine, vol. 24, no. 5, pp. 61-64, 2004. Dimitrios Hristu-Varsakelis and William S. Levine. “An undergraduate laboratory for networked digital control systems.” Control Systems Magazine, vol. 25, no. 1, pp. 60-62, 2005. Moallem, Mehrdad. “A laboratory testbed for embedded computer control.” IEEE Transactions on Education, vol. 47, issue 3, pp. 340-347, 2004. Edward A. Lee and Sanjit A. Seshia, Introduction to Embedded Systems, A Cyber-Physical Systems Approach, http://LeeSeshia.org, ISBN 978-0557-70857-4, 2011 Erik De Graaf and Anette Kolmos. “Characteristics of problem-based learning.” International Journal of Engineering Education, vol. 19, no. 5, pp. 657-662, 2003. Michael J. Prince, and Richard M. Felder. “Inductive teaching and learning methods: definitions, comparisons, and research bases.” Journal of Engineering Education, vol. 95, no. 2, pp. 123-138, 2006.
[2] [3]
[4]
[5]
[6]
Fig. 3.
A small solar panel with maximum output of 2.5 W.
[7]
D. Discussion The curriculum presented in this paper is ambitious, since the student population does not have as much experience with
3 http://www.ni.com/compactrio/ 4 http://www.quansercontrollabs.com/default.html
[8]
[9]
Patrick J. Martin
Electrical and Computer Engineering Program York College of Pennsylvania York, Pennsylvania 17403 Email: pjmartin@ycp.edu
Abstract—This paper presents an undergraduate controls curriculum that couples cyber-physical systems concepts with modern control design. The curriculum introduces cyber-physical systems topics to electrical and mechanical engineers that generally have less background in computing topics, such as concurrency and networking. This curriculum design increases undergraduate students’ awareness of cyber-physical systems.
• • • •
an ability to apply mathematical models of physical systems, cyber systems, and their composition an ability to design and conduct simulations and tests of a cyber-physical system and to analyze the results an ability to identify, formulate, and solve engineering problems that have both cyber and physical aspects a recognition for the need for, and ability to engage in life-long learning.
I.
I NTRODUCTION
Cyber-physical systems (CPS) require the application of many topics within modern controls research. Future controls engineers will be required to bring in additional concepts from communications and computing to solve these complex control problems [1]. There is a rising need to teach engineering students about the complexity introduced into modern control systems that utilize embedded computers and networks. This paper presents a new controls curriculum for undergraduates that incorporates several important topics from CPS with modern control design. At the undergraduate level, control systems is often stereotyped as a math-heavy, abstract subject. Consequently, [2] called for more integration of laboratory work into the controls curriculum. Since this work, several authors have investigated making controls more hands-on with laboratory work (e.g. [3], [4]). These lab activities give students valuable experience in the implementation of controllers on real systems. Many modern controllers are implemented in embedded computers, which requires students to know more about computing constraints, such as computational time and network delay. The work of [5] and [6] experimented with giving undergraduates experience in these topics, which are relevant topics in CPS education. These approaches create good lab experiences for students; however they should be accompanied by a course that blends the cyber and physical domains more readily. One example of a CPS focused curriculum is the work in [7]. This text provides a structure for introducing control principles to students in the computational sciences. However, there is still a need for the converse perspective: introducing computing and communications principles to students in the physical sciences and engineering. This paper presents an interdisciplinary, undergraduate controls curriculum that targets some of the proposed CPS learning outcomes:
To approach these learning outcomes, the curriculum blends modern control design with concepts from hybrid systems, networking, and concurrency. Since the students come from different engineering backgrounds, this curriculum must balance between physical and cyber systems topics to minimize student overload. At York College, electrical and mechanical engineers can take a two course sequence in control theory: Automatic Control and Applied Control. Automatic Control is a standard first semester course in control theory that covers classical topics such as frequency-domain modeling and analysis techniques. Near the end of the course students learn about PID control and apply use it to set the position of a DC motor using analog electronics. The sequel course, Applied Control, in the past was taught as a digital control course. However, mechanical engineers do not have the same background in digital signal analysis or programming as the electrical engineers. The overhead of teaching digital signal analysis and low-level programming distracted students from focusing on the control design concepts and tools. More recently, model-based design of controllers has become more common in the workplace. Products from Mathworks and National Instruments facilitate control design by providing tool chains that automatically generate low level code. Applied Control was re-designed for Spring 2013 to use the curriculum presented in this paper. All students in the course are familiar with feedback control; however, their knowledge of cyber systems varies. Thus, the course aims to introduce the cyber topics through project work and lectures. Students that take this course will be able to operate as control engineers and, furthermore, have a better perspective on the interplay between cyber and physical domains.
II.
C URRICULUM S TRUCTURE
B. Computing Models and Implementation The coupling of sensors, actuators, embedded computers, and networks presents many challenges for future control engineers. Students that do not have a strong background in computer science or engineering should study the cyber domain so that they can design high level control architectures. These architectures will need to take into account the constraints and abilities of cyber systems. For example, students interested in pursuing robotics need to be aware of discrete dynamics caused by computers switching among behaviors within an embedded operating system. At this point in the curriculum, students are introduced to finite state machines and hybrid systems. These topics develop their ability to apply mathematical models for cyber and physical systems. Interactive simulations of hybrid systems let students see the interesting effect hybrid control strategies have on physical dynamics. In model-based control design software, functions are represented as individual actors that process inputs and generate outputs. One advantage of this software design approach, is the ability to run concurrent control tasks with minimal thread locking issues. Students need to examine the principles of concurrency in actor and data-flow models so that their control implementations work across asynchronous computation elements. Exposing students to this theory is important for CPS education as more embedded systems are shipping with multicore processors. CPS engineers will be required to design systems that can balance power consumption, performance, and correctness. Thus, students should be aware of competing control loops in CPS applications. Consider the case where elements within a sensor network may need to conserve individual energy consumption; however, they may also be required to stream video data at a maximum rate. Recognizing physical and cyber system limitations in systems such as the example is an important skill when deploying CPS. This educational goal is best accomplished through an open-ended project that combines the foundational tools from control and computing. C. Project-Based Learning for CPS Most topics within this curriculum can be approached with lectures, in-class activities, and laboratories. However, CPS is a rapidly advancing field and students must synthesize new theories and technologies as they are developed. A projectbased learning activity (e.g. [8], [9]) complements this curriculum by providing a capstone experience for students to apply theory to solve a non-trivial problem. Project-based learning facilitates interdisciplinary team work, which is an important skill for students interested in developing CPS. Additionally, it provides motivation for students to reach outside of their academic comfort zone, which encourages their lifelong learning abilities. As an example of project-based learning for CPS, students in Applied Control at York College are assigned a project that utilizes the theoretical tools presented in II-A and II-B. The project involves the design and implementation of a multiagent solar farm. Each team of students is given a set of NI
An interdisciplinary controls curriculum for undergraduates needs to strike a balance between the students’ pre-requisite knowledge and the skills needed by the cyber-physical systems community. This particular curriculum builds student confidence in fundamental control system analysis and design topics before layering in the necessary CPS topics. Students apply their control design skills within simulation and hardware labs, which helps build intuition about state feedback and observers. They transition to CPS topics by studying finite state machines and hybrid systems, since they facilitate the modeling of cyber systems. Additionally, they learn about the basics of concurrency and networking and how they are used in modern control systems. A project-based learning activity is included so that students synthesize many of the topics within the curriculum into an application.
A. Foundations in Control Before students learn how to develop controllers for cyberphysical systems, they require a solid foundation in the core principles of modern control design. Students develop their intuition about physical systems by analyzing their models for stability. Additionally, students explore stabilization and regulation by applying classical PID control. PID control is a natural point of entry for the discussion of cyber systems topics, such as communication delay. To illustrate this topic, students explore the effects that delay has on DC motor performance when it is controlled over a network. They implement a PID controller on a laptop that issues control commands over a network interface to a real-time system that simulates a motor model. CPS are embedded within the physical world and are naturally multi-input, multi-output systems. Consequently, teaching state space analysis and design gives students a foundation for thinking about more complex system interaction. Students can apply controllability and observability to understand the structure of the system they are trying to control, and possibly re-design the system itself. These topics allow students to think about the actuation and sensing requirements of CPS applications. Students also learn about state feedback and its complementary topic of state observation for system regulation. Typically, these topics are reserved for graduate courses; however, introducing these tools in an undergraduate curriculum assists in developing intuition before invoking rigorous mathematical proof. Once students are familiar with these modern control techniques, they learn about optimal control through the Linear quadratic regulator (LQR) problem. Students experience these modern control design tools through LabView1 and Matlab2 simulation labs. Model-based design is emphasized so that students can focus on designing controllers. By following this foundational sequence, students become proficient at applying fundamental state-space analysis and design before adding on the computation and communication constraints in cyber systems.
1 http://www.ni.com/labview/ 2 http://www.mathworks.com/products/matlab/
Compact RIO computers3, one of which is shown in Figure 1. Each of these Compact RIO systems is interfaced to a Quanser SRV-02 servo motor4 , as shown in Figure 2.
advanced computing topics. However, undergraduate CPS education does not need to make every engineer an expert in these topics. Instead, it should make various types of undergraduate engineers aware of the field of CPS and the current challenges facing the research and industrial communities. Since this curriculum was deployed in the Spring 2013 semester, there is no assessment data to confirm if the curriculum structure satisfies the proposed outcomes. The author plans to take surveys at the conclusion of the course to measure student achievement with these learning outcomes. Some casual conversations with students indicate that their interest in the CPS-related topics is holding, despite the material being outside of their engineering specialty.
Fig. 1. A National Instruments Compact RIO computer with Quanser Q1USB module.
Additionally, the author will investigate the feasibility of moving some topics earlier in the undergraduate engineering program. A re-designed two-course sequence may be better suited to develop an undergraduate engineer’s appreciation for feedback control, computation, and communication. III. C ONCLUSION
Fig. 2.
A Quanser SRV-02 motor and its associated power amplifier.
The students must design a multi-agent control system where each agent actuates a solar panel (shown in Figure 3). The goal of the multi-agent system is to produce maximum total power generated by the solar panels while minimizing the control effort at each solar panel actuator. The agents operate independently and communicate over an ethernet network using TCP/IP. This project has multiple domain challenges for the students, from mechanical design and system modeling to algorithm selection and networked control. Furthermore, it highlights competing control design goals: energy efficiency and performance. To design their control system, the students need to perform research on current topics in networked and multi-agent systems as well as power systems.
This paper presented a controls curriculum that incorporates topics in cyber-physical systems. In addition to learning modern control design, students learn about concepts in computing that are outside of their major. Furthermore, they experience realistic physical and cyber constraints through a projectbased learning activity. This curriculum lays a foundation for undergraduate students to contribute to the CPS community in their future careers and research. ACKNOWLEDGMENT The author would like to thank the students of York College’s Applied Control course for participating in this experimental curriculum. R EFERENCES
[1] Ragunathan Raj Rajkumar, et al. “Cyber-physical systems: the next computing revolution” Proceedings of the 47th Design Automation Conference, ACM, 2010. Bernstein, Dennis S. “Enhancing undergraduate control education.” Control Systems Magazine, pp. 40-43, 1999. Arz´ n, K-E., Anders Blomdell, and Bj¨ rn Wittenmark. “Laboratories e o and real-time computing: integrating experiments into control courses.” Control Systems Magazine, vol. 25, no. 1, pp. 30-34, 2005. Bonnie S. Heck, N. Scott Clements, and Aldo A. Ferri. “A LEGO experiment for embedded control system design.” Control Systems Magazine, vol. 24, no. 5, pp. 61-64, 2004. Dimitrios Hristu-Varsakelis and William S. Levine. “An undergraduate laboratory for networked digital control systems.” Control Systems Magazine, vol. 25, no. 1, pp. 60-62, 2005. Moallem, Mehrdad. “A laboratory testbed for embedded computer control.” IEEE Transactions on Education, vol. 47, issue 3, pp. 340-347, 2004. Edward A. Lee and Sanjit A. Seshia, Introduction to Embedded Systems, A Cyber-Physical Systems Approach, http://LeeSeshia.org, ISBN 978-0557-70857-4, 2011 Erik De Graaf and Anette Kolmos. “Characteristics of problem-based learning.” International Journal of Engineering Education, vol. 19, no. 5, pp. 657-662, 2003. Michael J. Prince, and Richard M. Felder. “Inductive teaching and learning methods: definitions, comparisons, and research bases.” Journal of Engineering Education, vol. 95, no. 2, pp. 123-138, 2006.
[2] [3]
[4]
[5]
[6]
Fig. 3.
A small solar panel with maximum output of 2.5 W.
[7]
D. Discussion The curriculum presented in this paper is ambitious, since the student population does not have as much experience with
3 http://www.ni.com/compactrio/ 4 http://www.quansercontrollabs.com/default.html
[8]
[9]