engineering informatics

Engineering Informatics Education with Biologically Inspired Robots Rich Primerano David Wilkie William Regli Satyandra K. Gupta Department of Computer Science, Department of Electrical and Computer Engineering, Department of Mechanical Engineering and Mechanics, College of Engineering, Drexel University, Philadelphia, PA 19104, USA. {primerano, wilkie, regli}@drexel.edu Department of Mechanical Engineering, Institute for Systems Research, University of Maryland, College Park, MD 20742, USA. skgupta@eng.umd.edu Abstract—The design of robotic systems is a multidisciplinary effort requiring input from engineers and scientist in a variety of fields. Throughout the design process, numerous engineering models are developed to describe the various subsystems that make up a robotic device. We will present a biologically inspired robotic snake as a case study to demonstrate the typical engineering models that comprise a robotic system. Increasingly, simulation tools are being integrated early in the design flow to verify the operation of both individual subsystems and interconnections of multiple subsystems. Co-simulation often leads to difficulty because separate subsystems are often designed in incomparable software packages. I. I NTRODUCTION Engineering Informatics is the science of representation, simulation, archiving, and reuse of engineering knowledge. While other industries (e.g., financial, retail, digital entertainment) reap benefits and economies of scale from the information revolution, the manufacturing and engineering industries continue to lag behind [1]. Informatics, in general, deals with modeling, communication, visualization, interoperability, archival, and retrieval of information. By developing specialized curricula in the area of Engineering Informatics, we can better equip engineering students to tackle challenging problems in emerging engineering areas. Information technology is critical at every step of the engineering life cycle. This is especially true in emerging engineering areas, where students and researchers are continually building their own individual computational models and tools. This paper explores the use of biologically inspired robots in teaching multidisciplinary design and engineering informatics. II. P ROGRAM OVERVIEW AND BACKGROUND To focus the discussion of engineering informatics, we have chosen to use Bio-inspired robots as a case-study [2]. Terrain and applications that are not suitable for traditional robotic systems may prove suitable for bio-inspired robots. For example, search and rescue missions in complex urban environments require devices that can maneuver in collapsed buildings (e.g., cockroach), through ductwork (e.g., ant), and over a variety other obstacles. A snake-like robot may be used for planetary surface exploration, minimally invasive surgery, or inspection of piping and cabling. Such robots also have applications in homeland security and defense, enabling inspection of ships, containers and other structures too cramped for traditional robotic systems. There are a multitude of informatics challenges when designing devices such as these. Designing bio-inspired robots is a highly multidisciplinary activity. Components include sensors, actuators, structural elements, electronics, power source, and software. The actuated joints can create an extraordinary number of degrees of freedom (DOF), so physics-based modeling and simulation tools need to scale to handle vastly greater complexity. For example, design and simulation of a snake inspired robot leads to numerous problems: these devices can include hundreds of components, nearly all of which must interact for the snake to work. Additionally, one must capture engineering phenomena across all disciplines: mechanical, electrical, computer science, electronics, and environmental. The combination of scale and domain complexities demands extensive computation that requires highly optimized tools and efficient algorithms. Biologically-inspired robots seem to fascinate students from a wide variety of backgrounds [3] and appear to be a good candidate to not only hold student attention but even attract students to science and discovery. Abundant sources of information are available on how to design these robots [4], [5], [6] and they can be built as projects during the course of an academic semester or term. These sources are generally monodisciplinary, however, and are not, on their own, a suitable basis for informatics education. The challenge is that these robots are highly multi-disciplinary, consisting of sensors, actuators, mechanical structures, electronics, and software. Components from these diverse disciplines need to be tightly integrated to create successful robots. The challenge to the student roboticist is to develop a working understanding of each required disciplines, an understanding adequate enough to appreciate the complex issues that emerge at their interfaces. In this paper, we present work done at Drexel University and The University of Maryland toward developing a curriculum to teach engineering informatics. III. E DUCATIONAL ACTIVITIES AT D REXEL U NIVERSITY The initial course at Drexel was executed in the Fall quarter of 2006. This one seminar class had 12 students spanning computer science, mechanical engineering, computer engineering, software engineering and electrical engineering. Students were either senior-level undergraduates or graduate students; and most were computer science. The goal of the class was to introduce students to engineering informatics while building comprehensive engineering models of biologically-inspired robotic systems. Each student had to develop their own design of a bio-inspired locomotion mechanism, build it with a robot construction kit (most chose to use Lego Mindstorms kits) and implement a complete 3D model with kinematics simulation and, if possible, dynamics as well. Mindstorm robots were a popular choice because they allowed students with limited electromechanical system design experience to quickly build physical prototypes. During the class, student learned to: • identify problems resulting from the interdisciplinary; interactions in bio-inspired robots; • perform system engineering to design, test and build biorobots; • apply informatics principles to bio-robot design and testing; • use appropriate hardware and software tools for robot design/analysis; To model their robotic designs, students were given access to a variety of computational and engineering informatics tools, including MATLAB, SolidWorks, Pro/Engineer, ACIS, ADAMS, and MAPLE. Software libraries for rigid-body dynamics, such as the Open Dynamics EngineTM , were also made available to the students. While these libraries, most often used in computer gaming applications, lack the accuracy of engineering simulators, they have other advantages: (1) they have been designed to perform real time simulation on single computers; (2) they are useful for making qualitative predictions about the behaviors of physical systems, as demonstrated in [?] to find a fast method of movement for snake-like robots; (3) most importantly, these libraries offer a full view of the computational process of simulation and require the students to consider needed geometric transformations, controller interaction, and generally motivate students to consider the complexity of the data involved. During the course, students documented their output, designs, design rationale and simulations in weekly reports archived on a couse wiki. At the end of the project, demonstrations were given of the completed locomotion platforms as well as of the simulation models developed for them. Figure 1 shows an example of a CAD model that was developed for a walking robot designed in the course. A. Design Example As a case-study for the discussion of engineering informatics and multidisciplinary design, we will use a robotic snake that has been designed at Drexel University [2]. The robot, Fig. 1. A walking robot designed in Drexel University’s course. shown in Figure 2, is designed to move forward by crawling with many small feet operating in synchronization. Along its length are fifteen motors with eight processors to control them. The assembled robot contains over one hundred parts, and is driven by the operator through a computer interface. Fig. 2. A biologically inspired robotic snake. The snake provides an interesting example because it contains elements of electrical, mechanical, and computer engineering, computer science, and biology. Through the course of the robot’s design, numerous engineering models were developed to describe the operation of the snake’s many subsystems. B. Engineering Models Complex electromechanical systems such as biologically inspired robots offer a unique educational opportunity. The development of such systems is often a group effort. Engineers from multiple disciplines must collaborate, and they generally require a familiarity with areas of engineering outside their own. Biologically inspired robots provide an interesting and effective means of teaching the basics of multidisciplinary design. In our robotic snake case study, we have developed data sets representing the following design elements: • Solid models of mechanical components • Finite element analysis of mechanism • Kinematic model of robot • Kinematic simulation of mechanisms • Dynamic simulation of mechanisms • Schematics of electrical hardware • Board layouts of electrical hardware • Control and Communication firmware • Control algorithm simulations • Controller/Robot co-simulation • Path planning algorithms Communication System Mechanism Design Kinematics Circuit Design Mechanical Simulation Electrical Controls Finite Element Analysis Dynamic Simulation Kinematic Simulation Ground Hugging 1 Fig. 3. Stability Monitoring Sensor Processing Typical elements of electromechanical system design. These engineering models serve two purposes during design and construction of the robot. Naturally, they are used at the end of the design process to implement the physical robot. More importantly, however, they can also be used throughout the design process to simulate the operation of the robot. Figure 3 depicts the multidisciplinary nature of robotics design and gives a sense of the data representations that comprise a typical robotic system. During simulation, software tools that model the robot’s kinematics, controller, and dynamics must communicate with one another. C. Challenges At several stages throughout the design process, simulation tools were used to verify the proper operation of mechanical, electrical, and control systems. The main issue encountered in this project was the incompatibility of data representations between the design tools and the simulation tools. For example, solid model files contain much more information than just geometry data. Parametric relationships, datums, part-topart constraints, and other information contained in the solid model are sometimes lost when the geometry is imported into a dynamic simulation tool. As a result, designers who wish to perform design, simulation and redesign iterations must take intermediate steps to ensure that all required data is carried from one tool to another. At the time of simulation, multiple engineering models must be integrated into a format compatible with the simulator used. For example, the CAD model of a mechanism needs to be integrated with a kinematic model and a controller during dynamic simulation. The kinematic model may have been created in the CAD system, but be incomprehensible to the simulator (if, for example, the simulator uses a mesh representation and the kinematic model is defined in terms of solid model features). Students most successful at simulating their designs bypassed CAD, creating their designs directly in the simulator. They avoided the integration issues that arise in moving from design tools to simulation tools. This solution does not introduce students to many of the issues that arise at the interface between CAD and simulation, however. This is important because it is often the case that CAD models already exist in a particular format. A more desirable solution is to have an established path for students to follow in order to create and integrate their models. This would include a tool or choice of tools for each type of model and a plan to import and integrate these models in the final simulation. For the simulation of the snake-like robot, two approaches at creating a comprehensive simulation were tested, one using a commercial simulator and one using an open-source rigidbody dynamics library. The commercial simulator was able to load a CAD model, but some information was lost in the process. The kinematic model had to be created inside the simulator using a feature-based method. A controller was modeled in a different commercial tool that was able to connect to the simulator. The rigid-body dynamics library could only use mesh files derived from the CAD model. The kinematic model had to be created in terms of axes and points, which proved very difficult to obtain from the CAD program (a separate program had to be written to extract the axes of features for the constraint definitions). A controller was written in code within the program. D. Course Results The bio-inspired robotics theme proved to be very compelling to the students taking this course. Initial weeks were spend exploring different kinds of locomotion mechanisms, with most students ending up settling on some kind of walking platform. Many students chose to build the platform first then create the informatics models, the opposite approach than might be taken in industry or in practice. The modeling process took the majority of the course time, on the order of four to five weeks of a ten week quarter. The reasons for this stemmed from the exploration of a wide variety of tools and approaches to modeling mechanisms, kinematics and dynamics associated with the locomotion mechanisms. Students were given crash courses on CAD solid modeling and then taught themselves packages such as Pro/Engineer, SolidWorks, MicroStation, I-DEAS, ADAMS and Blender. Ultimately, a combination of SolidWorks and ADAMS proved to be the best fit for many students. One unexpected and very positive result of this activity was that students who never worked with physics-based simulation or CAD packages taught themselves what they needed to know to accomplish the assignment. Many of these student had never touched a CAD system previously yet managed to be producing parametric assemblies with constraints and joint information in a matter of two weeks. IV. E DUCATIONAL ACTIVITIES AT THE U NIVERSITY OF M ARYLAND The course offered at the University of Maryland also centered on biologically inspired robotics. Due to logistics challenges, we decided to only use the software tools that were available and familiar to the students taking the course. These tools were (1) Pro/Engineer software for design, kinematic simulation, and finite element analysis, (2) Matlab for controller design and dynamics simulations, and (3) Microbasic for micro controller programming. The purpose of new course was to: • Assess effectiveness of this course in attracting students’ attention to focus on the engineering informatics challenges. • Identify the tools that will be needed to teach engineering informatics principles. • Identify pedagogical approaches effective for teaching engineering informatics. • Assess effectiveness of engineering informatics principles in successfully realizing multi-disciplinary products. This course covers the fundamentals and applications of biologically inspired robots. There are abundant sources of information on how to design these robots, and they can be easily built in semester long projects. These robots are highly multidisciplinary, consisting of sensors, actuators, mechanical structures, electronics, and software. Components from these diverse disciplines need to be tightly integrated to create successful robots. Thus, the course was designed to convey this multidisciplinary knowledge in three main parts: • Fundamentals of Traditional Robotic Manipulators: One must be familiar with the fundamentals of traditional robots in order to conceive, analyze, and create new robot designs. This part of the course begins with the history and taxonomy of traditional robots. Different popular robot configurations are introduced. This part also covers forward kinematics, inverse kinematics, and dynamics of serial manipulators to analyze proposed robot designs. • Fundamentals of Biologically Inspired Robotics: This part of the course begins with a discussion on the role of biological inspiration in robot design. Several examples of bio-inspired robots are discussed in detail, including the motivation and biological inspiration for their design, as well as technical specifications and comparisons to conventional robots. • Design and Fabrication of Biologically Inspired Robots: This part of the course covers techniques for designing and fabricating biologically inspired robots, as well as selecting and programming micro controllers for controlling biologically inspired robots. The basics of numerically controlled machining and injection molding technologies for fabricating structural members of biologically inspired robots are also described. During the course, three person groups were tasked to design and build biologically-inspired mobile robots. Figure 4 shows an example of a crocodile inspired robot and figure 5 shows an ant inspired robot. Each was designed during the University of Maryland course. Wheeled robot designs were not allowed, and robot had to have on-board power supply and electronics. Tethered-designed were discouraged. The design goal was to build a robot that will go along a straight line for two minutes. The robot performance will be measured in terms of the length traveled expressed as body lengths. To ensure robustness of the gait design, the robots were tested on two different surfaces; parking lot and carpet. Teams were given a servo motor and a compatible microcontroller recommendation. This neutralized any advantage a team can have by using stronger servo motors. As a part of their project, students had to: Project #2 Example: Croco Fig. 4. A crocodile inspired robot designed during the University of Maryland class. • • • • Design and build the mechanical structure to realize an approximation of a biological creature Design a gait Implement the gait in the software Design the necessary circuit to connect the servo motors to the micro controller ect #2 Example: Ant Robot • Select the appropriate batteries Fig. 5. class. An ant inspired robot designed during the University of Maryland Despite extensive use of CAD and simulation tools, all teams went through a very large number of design iterations. Common reasons behind these iterations were as following: • Motor torque was not sufficient to drive the designed 9 structure. There was no convenient way for teams to get the robot pose data from Pro/E to Matlab. Hence, during the gait there existing poses that required large torque. • There was excessive deflection in the robot limbs (robots limbs were made out of lightweight plastics to reduce weight and attain high speed) and it interfered with the intended gait. There was no convenient way for teams to accurately estimate limb deflections and incorporate them in the kinematic simulations). • Robots made awkward contacts with the floor due to slight errors in the synchronization of different motors. Hence the gait did not function correctly. • Friction was not estimated and modeled correctly. Hence some robot designs showed excessive slipping. • There was no convenient and accurate way to estimate energy consumption. Hence teams had to use extra batteries and as a result increased the mass of the robots. • They were unable to account for manufacturing tolerances. Hence, considerable time was spent on tailoring joint angles to compensate for manufacturing tolerances. • Student teams often had trouble in controlling the software versions. Often, they would put the wrong version of some piece of code on the robot. They also did not document their rationale well and sometimes had trouble in tracing the reason for a design change. We believe that many of these problems and associated design iterations can be eliminated by use of engineering informatics principles. By educating students in the these principles, we hope that they will be able to make better use of simulation tools to enable minimal reliance in hardware prototyping. A. Course Results There was an overwhelming response to this course and a very large number of students signed up for the course. Students showed enthusiasm through the duration of the course, confirming that biologically-inspired robotics appears to be a good choice for attracting students’ attention and can be used to emphasize engineering informatics challenges. We believe that new tools will need to be developed to properly introduce engineering informatics principles into the curriculum. These tools will (1) provide tighter integration across domains, (2) allow specifications of gait designs, (3) provide better visualization of robot movements, and (4) manage engineering models. It appears that a good pedagogical approach to teaching engineering informatics principles will be just-in-time lecturing. The idea behind this approach will be to assign projects such that students who neglect the informatics principles severely limit their chances of completing the project in time. In such situations, students will be motivated to learn and use the informatics principles. Additionally, a significant portion of the project grade rewards the use of informatics principles. Our initial experiences indicate just using the best-in-class CAD, simulation, and analysis tools from specific disciplines cannot guarantee success in design projects. We believe that without applying the engineering informatics principles, it will be impossible to realize successful products in a timely manner. V. C ONCLUSION Several important results emerged as a result of the educational activities at Drexel University and the University of Maryland. Educational Results Biologically inspired robotics projects appear to be effective at introducing students to engineering informatics and multidisciplinary design. In our experience, these projects have captured the interest of science and engineering students and have given them an appreciation of the aspects of robotics design. Students with little or no prior experience using design and simulation tools were able to quickly teach themselves the basics of these packages. Despite having access to numerous commercial and opensource software packages, however, students still had difficulty producing system-level simulations that combine dynamics, kinematics, and control algorithms. Implementation Results The intent of these courses is to show students how software tools can be used to simplify the design process with virtual prototyping. We have noticed that while students often learn the basics of particular CAD software tools rather quickly, they don’t take full advantage of capabilities that these packages provide. Often, we see students build a physical prototype, then attempt to construct a simulation model that reflects its behavior. One reason for this may be the difficulty encountered in interfacing the engineering models describing different subsystems into a system level simulation. Another reason for this apparently backwards use of technology may be that students specializing on one aspect of robotics design (e.g. mechanical engineering) may be unfamiliar with other aspects of design (e.g. electrical and software design). We hope that through continued improvement in our engineering informatics curriculum, we can enable engineers to more effectively make use of simulation tools throughout the design process. ACKNOWLEDGMENT This work was supported in part by National Science Foundation (NSF) Office of Cyber-Infrastructure Grants OCI0636235 and SCI-0537370. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation or the other supporting government and corporate organizations. R EFERENCES [1] W. Regli, “The need for a science of engineering informatics,” AI-EDAM, vol. 21, pp. 23–26, 2007. [2] D. Wilkie, R. Primerano, and W. Regli, “A simulation-centric framework for design of a robotic system,” ASME IDETCE, pp. 23–26, Aug. 2007. [3] H. A. Bruck, A. L. Gershon, I. Golden, S. K. Gupta, J. L.S. Gyger, E. B. Magrab, and B. W. Spranklin, “Training mechanical engineering students to utilize biological inspiration during product development,” Bionispiration and Biomimetics, 2:S198-S209, 2007, 2007. [4] K. Williams, Amphibionics : Build Your Own Biologically Inspired Reptilian Robot. McGraw-Hill/TAB Electronics, 2003. [5] G. A. Bekey, Autonomous Robots: From Biological Inspiration to Implementation and Control. The MIT Press, 2005. [6] S. Hirose, Biologically Inspired Robots: Snake-like Locomotors and Manipulators. The MIT Press, 1993.