Transcription
To be presented at:IEEE Aerospace Conference, Big Sky, Montana, Mar. 7-14, 2015On the Need for Significant Reform in UniversityEducation, especially in Aerospace EngineeringLyle N. LongThe Pennsylvania State University233 Hammond BuildingUniversity Park, PA 16802814-865-1172lnl@psu.edunever use, and conversely, they are not being taught materialthat will be critical in their future career. While in somefields it is important to teach the detailed history of thesubject, in engineering the students need to learn timely,important, and relevant material. Much of what is beingtaught is taught for historical purposes. This is not fair toour students. They need to learn how to create systems forthe information age, not learn how to build systems from 50years ago.Abstract—This paper discusses the need for academia to bemore responsive to the needs of students and society. The slowspeed of change in academia is causing our educationalprograms to lose value. Technology has been advancing at anexponential rate for decades, yet academia changes at almostglacial speed. While the example used herein (engineering, andin particular aerospace engineering) is one familiar to theauthor, the ideas in this paper may well apply to all academicdisciplines, including Science and Liberal Arts. The problemsare due to rapid changes in technology, inflated bureaucraciesat universities, the emphasis on revenue and research, andlimits to human learning. There is no question that this is theInformation Age, yet academia has not adjusted to thisdramatic new world. Students need to understand this and bepro-active to prepare themselves for the future and choose theright courses, majors, and universities.If aerospace engineering education does not change, it willbecome as irrelevant as railroad engineering, which wouldbe terrible. Aerospace systems are crucial to the U.S. Theyrepresent significant export dollars, they are crucial to U.S.defense, they are crucial for transportation, and they arecrucial to exploring the universe.TABLE OF CONTENTSThe issue here is not about students getting jobs. There is ashortage of engineers, and so recent graduates can usuallyget jobs. The real issue is preparing them for the 21stcentury. For nearly 10 years the author has been surveyingjob openings at the large aerospace companies (e.g. Boeing,Lockheed-Martin, and Raytheon). The results have beenfairly consistent for 10 years. Table 1, presents the numberof job openings at Lockheed-Martin in 2011, and is a typicalrepresentation of the needs of other companies as well.1. INTRODUCTION .12. EXPONENTIAL CHANGES IN TECHNOLOGY .23. COMPUTER PROGRAMMING .24. AEROSPACE ENGINEERING CAREERS .35. AEROSPACE ENGINEERING CURRICULA .46. CONCLUSIONS .6REFERENCES .7BIOGRAPHY .7NewGraduatesExperiencedEngineersSoftware Engineering45172Systems Engineering42342Information Technology251041Mechanical Engineering642Electrical Engineering348Aerospace Engineering1291. INTRODUCTIONIt has been about 7 years since the paper “The Critical Needfor Software Engineering Education” [1] was published. Inthat paper (which should be read before this one) it wasexplained why we need many more people educated inprofessional software development. It also discusses whysoftware is crucially important.In addition, it has been 11 years since the Journal ofAerospace Information Systems [3] was created, which isrecognition from the AIAA that computing, information,and communication is a key pillar of aerospace engineering[2]. Unfortunately, there have been few changes inengineering education during this period. In this paper thecase will be made more forcefully (and bluntly) that ourtechnological future is not being well served by currentuniversities. Students are being required to learn anenormous amount of very difficult material that they willTable 1. Number of job openings at Lockheed-Martin in2011 [9].978-1-4799-5380-6/15/ 31.00 2015 IEEE1
Aerospace engineering education programs are clearly notkeeping pace with industry if industry does not really needaerospace engineers. They mainly need software engineers,system engineers, and information technology experts.Over the course of a career, the mismatch between existingknowledge and human understanding grows exponentially.It is important for faculty, students, and employers toappreciate this. We are in the information age, not theindustrial age. What professors learned 30 years ago mightnot be important to current students. And it is extremelydifficult for humans to keep up with technology.The other problem is that making changes in academia isincredibly difficult and slow, especially in large researchuniversities. Any change to the curriculum has to beapproved by the departmental faculty, department Head,other departments, the college, the Dean, the faculty senate,the graduate school (for graduate courses), and the Board ofTrustees. This process can take several years, which meansit will be obsolete before it is approved. Universities need tobe able to react more quickly.3. COMPUTER PROGRAMMINGSince technology is changing so rapidly, some people do noteven appreciate what they do not know. There are stilldebates in academia about whether we should teachengineers C or Fortran, even though Fortran is a very oldand out-of-date language. The Air Force even states itshould not be used [6]. But this discussion illustrates howlittle people know about modern computing and how out oftouch they can be.Figure 1. Comparison of linear learning of humans withannual information doubling rate of technology.The IEEE recently evaluated programming languages [7],and there were 23 computer languages that are moreimportant than Fortran. The top four languages were: Java,C, C , and Python. They evaluated languages in fourdifferent ways (i.e. Overall, Job availability, Which aretrending, and Which are heavily used for open systems). Inall four cases, Java, C, C , and Python were the mostimportant languages. They also evaluated the languages infour different application areas (Web, Mobile, Enterprise,and Embedded systems). All four of the above languagesare heavily used in enterprise computing. Java and Pythonare heavily used for web applications. Java, C, and C areheavily used for mobile devices. And finally, C and C (and assembly language) are the key languages forembedded devices. Fortran is no longer important and ourstudents should not be taught it as their first language, yetsome academics still debate it’s relevance. Incidentally,Fortran is very simple and the students could easily learn iton their own later on if they need it, and Matlab is muchmore useful than Fortran. The people that ask whether weshould be teaching C or Fortran don’t know what theydon’t know. The world has moved orders of magnitudeaway from that question. A good way to assess yourprogramming abilities is to determine where you fit in theprogrammer competency matrix by Joseph [8]. Manytraditional aerospace engineers (including faculty) wouldnot pass the first level.2. EXPONENTIAL CHANGES IN TECHNOLOGYTechnology has been evolving at an exponential rate forhundreds or thousands of years. This has been welldescribed by Kurzweil [4], who discusses the implicationsof these rapid changes. It is now changing so fast that it isquite difficult for humans to keep up with it. One of theproblems is that human learning is linear, while technologychanges exponentially. According to Gantz and Reinsel [5](and others) the amount of online information is doublingevery year. In 2020 it is expected that there will be 1022Bytes of information online. The human brain has about1015 synapses and each one can store roughly 1 Byte, so theonline data will be roughly 107 times larger than a humanbrain can hold. Meanwhile, academia has changed little in500 years.The implications of this can be illustrated with a very simple“back of the envelope” calculation. Imagine a person readsthree books a week and remembers all this information(which is clearly not possible but let’s consider it as anupper bound). An average book is about 400,000 Bytes,which gives about 60 Mbytes/year if someone can read andremember the information in three books per week. Figure 1shows the rate of information growth (doubling every year)compared to the rate of human learning (assuming 3 booksper week), where both curves start at unity. This is an oversimplified example, of course, but it helps illustrate themismatch between human learning and technology change.2
4. AEROSPACE ENGINEERING umentation, data collection, computers, and software.I always encourage students to do a job search early in theireducation. Too often they will get excited about some aspectof engineering, and then when they graduate they find thereare no jobs in that area. My 2008 Crosstalk paper [1]discusses which skills and knowledge major aerospacecompanies (and government labs) are looking for, and it isnot the traditional areas of aerospace engineering. Theyshould also search the websites of large and small aerospacecompanies [9-12]. They should also read the Broad AgencyAnnouncements (BAA) from the major governmentagencies [13 - 16], which will show them which researchareas are important to the DOD, NASA, etc. (and they arenot traditional aerospace engineering topics).To put this in perspective, the next generation of long-rangebombers will cost more than 500 million each, and 50-70%of this cost will be in computers and software, which istypical of modern aerospace systems. Yet in a typicalAerospace Engineering program, only about 5% of thecurriculum is devoted to computers and software. Thereforethe students cannot appreciate or even contribute to theentire system. They may not even be able to communicatetechnically with the majority of the other engineers. IfAerospace Engineering is not modernized, it will becomeirrelevant. We should not be training engineers how todesign and build aircraft from the 1950’s.ABET [17] is another issue. In order for an educationalprogram to be accredited, ABET requires extensive effort bycolleges, which is often of limited value. For aerospaceengineering programs, ABET [18] states:Software has been called the Achilles’ heel of aerospaceengineering [19], but yet it is hardly taught in aerospaceengineering programs. The Boeing 777 has more than 1000processors onboard and nearly 20 million lines of code.Most of the problems that occur on new aircraft andspacecraft are software related, yet software engineering isnot normally part of aerospace engineering.“Aeronautical engineering programs must preparegraduates to have a knowledge of aerodynamics,aerospace materials, structures, propulsion, flightmechanics, and stability and control.”Also, one of the most rapidly growing areas in aerospace isdrone technology (or UAV’s). It is very easy to buyunmanned radio-controlled aircraft with efficient enginesand propellers [20]. You can also buy autopilots off theshelf [21]. Therefore, anyone can buy a drone and theaerodynamics, structure, propulsion, and control system isalready done. Again, all the “traditional” aerospacedisciplines are mature. The difficult aspects of drones aremainly in computers, software, sensors, signal processing,artificial intelligence, networking, navigation, etc.; almostnone of which is being taught, especially to aerospaceengineering students. When the author first started workingon UAV’s (about 10 years ago) he was shocked to realizethat most of the aerospace engineering students were of littlevalue, since we needed people who understood software,sensors, numerical methods, and computers. We could buythe aircraft, engines, and autopilots.These are the “traditional” technology pillars [2] ofaerospace engineering. Unfortunately, this is how aerospaceengineering was described 50 years ago, when aircraft hadno onboard computers, networks, or data. Students shouldbe able to design and build the aerospace systems of thefuture, not reproduce ancient technology. Today, withoutdetailed knowledge of computers, software, networking,embedded systems, and software and systems engineering;the above curricula would not be of much value. The ABETdescription mentions nothing about aerospace engineerslearning about computers or software. This is ridiculous. Allthe above topics are very mature compared to computersand software, and few jobs exist in these areas. Fortunatelyfor students, there is a big need for engineers, andemployers often train new employees in areas they are notwell versed in. So the whole complicated and extremelytime consuming ABET process for aerospace programs isbased upon erroneous assumptions.In addition, some of the key problems in designing UAV’sare completely ignored in most aerospace programs: faulttolerance, single-point failure analysis, command andcontrol, radio frequency (RF) links, and compatibility withNext-Gen systems.In addition, if someone wanted to work in one of these areas(aerodynamics, aerospace materials, structures, propulsion,flight mechanics, and stability and control), they would haveto know a great deal about computers also. Aerodynamicsand propulsion work is now performed on massivelyparallel Linux-based computers using computational fluiddynamics, which is base upon numerical methods andmathematics. Likewise, the behavior of materials andstructures are determined using finite element analyses onLinux-based parallel computers. Stability and controlrequires extensive understanding of mathematics, andusually requires the use of Matlab and Simulink. So todayeven the original pillars of aerospace engineering are builtupon computers and software. Even experimentalSimilar stories could probably be presented for most otherdisciplines (engineering, science, liberal arts, etc.). We needdramatic and significant changes in academic curricula;incremental changes amidst exponentially fast technologychange will not work. Our students deserve better. Allstudents (no matter what their major is) would be wise toobtain a Minor related to computing and software.Often the students understand these issues better than thefaculty. One institution (which will go nameless here)started offering an undergraduate minor in computerscience. It was so popular that they could not handle the3
number of students and cancelled the Minor program.Obviously academia is not run like a traditional businessThe University should have put more resources into theprogram to keep it alive.majors) would be well served by obtaining a minor incomputing and/or software. In addition, most computerscience graduates are not educated in software engineering.The graduate and undergraduate minors have been effective,but too often students are being asked to learn material intheir major that they will never use. In order to spark adebate about this, this paper will discuss the curriculum inmore detail, and suggest updates.5. AEROSPACE ENGINEERING CURRICULAIn my experience it is almost impossible to change academiccurricula. Too many people have vested interest in keepingthe status quo, and there are few financial incentives to do it.There is a way around the impasses however, by creatingundergraduate and graduate Minor degrees. These allowstudents to get the educational components they need, andthey get credit for them. They are also much easier to createthan it is to modify existing Majors, however they are notusually well funded.2 shows the current courses required at Penn State for adegree in aerospace engineering, which is quite similar toother programs of this kind around the world. The numbersindicate credit hour requirements. A 3 credit course meetsthree times a week for 50 minutes each. The degree requires131 credits in total. They need to take 31 credits in theLiberal Arts, which is important (and also a rigidrequirement). They need 20 credits of Math, which includescalculus, ODE’s, PDE’s, linear algebra, complex variables,etc; and 10 credits in physics. They do not have to take anycourse in Systems Engineering, and they are only requiredto take 3-6 credits in computing and programming. Thetraditional (and mature) topics (aircraft or spacecraft design,aerodynamics, propulsion, stability & control, lab, andstructures) require 52 credits. The remaining 12 credits aretechnical electives. This program is far too light incomputing and software, and too heavy in the traditionalareas. It also gives the students a false impression of what isimportant in the 21st century.In about 1999 the Penn State Graduate Minor in HighPerformance Computing was created [22]. This was toaddress the need for scientists and engineers to learnnumerical methods and parallel programming. Since then ithas been renamed the Graduate Minor in ComputationalScience. Information on this program, courses, students, etc.can be found on the webpage [22]. We have graduated 248graduate students, who are now working around the world,and we currently have 86 more enrolled. These studentscome from across the university, not just engineering. Onestudent told me that the Minor was more important to hercareer than her PhD major (chemical engineering). AnotherPh.D. student said sample codes from my courses are nowpart of weather forecasting codes. A Ph.D. student inastrophysics said my course was the most useful one he hadhad in graduate school.The courses required for the Aerospace Major combinedwith the Minor in IST are shown in Figure 3. At a minimumthey need 10 additional credits (for a total of 141), but inthis approach they take 22 total credits in computing-relatedcourses. And if they schedule their courses carefully theycan still graduate in four years (many of the IST courses areoffered in the summer and/or online). They still have thesame number of credits in the traditional aerospace courses,liberal arts, math, and physics. And they still have 6 creditsavailable for electives, which they should choose carefully(i.e. they should choose more courses in areas such ascomputing, numerical methods, control theory, robotics, ormathematics).Building upon the success of the Graduate Minor, in 2006an undergraduate Minor in Information Science andTechnology (IST) for Aerospace Engineers was created.This is described on a webpage also [23]. Undergraduatesneed to complete 19 credits in computing, networking,information, and software to receive this minor. But it isrelatively easy for them to obtain this Minor, since some ofthe courses (about 9 credits) can be counted towards themajor as well. Two new courses (Advanced ComputerProgramming [24] and Software Engineering [25]) werealso created that can be used towards this Minor. Again, thestudents usually know what is important. These courses arealways full, and sometimes there is a waiting list. Each yearthere are 50-60 students in these courses. Undergraduatestudents have told me these courses helped them in gettinggood jobs. Employers are usually very interested inaerospace engineering students who also know aboutcomputing and software.Figure 4 shows a new recommended curricula for aerospaceengineering, just for discussion purposes. It assumes thenumber of total credits must remain at 131 (this is typicallyvery difficult to change at a university). It is unlikely,unfortunately, that this would ever get approved by anexisting aerospace engineering department. In this curriculathere are 21 required credits in computing and software, and3 more credits in math (for numerical methods). Thetraditional areas of design, structures, and fluids have beenreduced by 6, 3, and 3 credits, respectively. The technicalelectives have been reduced as well. In addition, a course inSystems Engineering would be required. These studentswould have a tremendous foundation on which to build theircareers in the 21st century and would be highly sought afterby industry, government labs, an graduate schools. Inaddition, they could easily complete this in four years.Of course there are many undergraduate and graduateminors offered at most universities. Every student shouldobtain at least one minor. Too often they obtain minors thatare too similar to their major however, which is a mistake.Some excellent minors students should consider aremathematics, statistics, business, computer science,information technology, etc. But every student (in all4
Figure 2. Breakdown of credits required for Aerospace Engineering degree at Penn State University.Figure 3. Breakdown of credits required for Aerospace Engineering degree with IST Minor at Penn State University.Figure 4. Breakdown of credits required for recommended Aerospace Engineering degree.5
Figure 5. Comparison of the three different curricula (current, current plus IST Minor, and recommended).Figure 5 summarizes the three different curricula, andmakes it easer to see the key differences.This is a long list of topics (and not complete). The first tenitems should be included in any undergraduate program inaerospace engineering. The first six items should be learnedby almost all students in any major. If they do not, then thestudents need to find a way to learn them on their own(unfortunately). Many of the other ones could beincorporated into M.S. or Ph.D. programs (which we havedone in our Graduate Minor program [22]). The studentswould be much better able to tackle these in graduate schoolif they have a 21st century undergraduate program to buildupon. All undergraduate students should obtain some sort ofMinor in computing or software. It is not recommended thatstudents Major in software engineering, it is better as aMinor or as a graduate degree.In most of this paper reference has been made to the topic“computers and software.” It is important to describe this inmore detail. Some of the topics that would be incrediblyvaluable for students (and their future employers) would be:Undergraduate:1.2.3.4.5.C/C Web programming (Java and HTML)PythonMatlab & SimulinkLinux (including real-time operatingsystems, RTOS)6. Numerical methods and discrete math7. Electrical Circuits and Microprocessors8. Robotics or UAV’s9. Software Engineering10. Systems Engineering6. CONCLUSIONSComputers and software are crucial parts of aerospacesystems, and the largest and most important part. Too oftentraditional aerospace engineers (including faculty) limittheir consideration of computing and software as tools, andnot part of the system. Faculty are often poorly trained incomputing and software, since the technology changes sorapidly. Academic programs (in all disciplines) are veryoften badly out of date because they are difficult to change,but if the U.S. is to remain preeminent in engineering andscience we must modernize our curricula. This paper hasmainly addressed undergraduate education, but similarissues exist in graduate programs as well. Technology hasbeen changing at an exponential rate for a thousand years,while academia has changed little in 500 years. Studentsneed to understand this and take charge of their education. IfUniversities do not modernize their curricula, students mustlook for appropriate Minors (e.g. [22] and [23]) which willgive them 21st century skills.Graduate:1.2.3.4.5.6.7.8.9.10.Advanced programmingNetworksEmbedded devicesIntelligent Systems /Artificial IntelligenceNeural networksNavigation (including GPS)AvionicsElectromagneticsParallel computingStatistics and Big Data6
REFERENCES[21] http://www.cloudcaptech.com/[1] Long, Lyle N., " The Critical Need for SoftwareEngineering Education," in CrossTalk: The Journal ofDefense Software Engineering, Vol. 21, No. 1, Jan, 2008.[22] http://www.csci.psu.edu/[2] Long, L.N., "Computing, Information, andCommunication: The Fifth Pillar of AerospaceEngineering," Editorial, Journal of Aerospace Computing,Information, and Communication, Vol. 1, No. 1, Jan., 2004.[24] http://www.personal.psu.edu/lnl/424pub/[23] http://www.personal.psu.edu/lnl/ist/[25] http://www.personal.psu.edu/lnl/440pub/[3] AIAA JAIS Journal, http://arc.aiaa.org/loi/jaisBIOGRAPHY[4] Kurzweil, R., “The Singularity Is Near: When HumansTranscend Biology, Penquin Books, 2006.Lyle N. Long received a B.M.E.degree from the University ofMinnesota, Minneapolis, MN,the M.S. degree from StanfordUniversity, Stanford, CA, andthe D.Sc. degree from GeorgeWashingtonUniversity,Washington, DC. He is aDistinguished Professor ofaerospaceengineeringcomputational science, neuroscience, and mathematics atThe Pennsylvania State University, State College, PA,USA. He is the Founder and Director of the GraduateMinor Program in Computational Science. He was theFounding Editor-in-Chief of the AIAA Journal ofAerospace Information Systems. From 2007 to 2008, hewas a Moore Distinguished Scholar with the CaliforniaInstitute of Technology, Pasadena, CA, USA. He waspreviously with Lockheed Aircraft (Burbank, CA) andThinking Machines Corporation (Cambridge, MA). Hehas authored more than 250 journal and conferencepapers. His current research interests include neuralnetworks, software engineering, cognitive robotics, highperformance computing, and unmanned vehicles. Dr.Long received the Penn State Engineering SocietyOutstanding Research Award in 1996, the 1993 IEEEComputer Society Gordon Bell Prize for achievinghighest performance on a parallel computer, and theLockheed Aeronautical Systems Company Award forexcellence in research and development. He is a Fellowof the American Physical Society and the AmericanInstitute of Aeronautics and Astronautics.[5] Gantz, J. and Reinsel, D., “The digital universe in 2020:big data, bigger digital shadows, and biggest growth in theFar East,” Dec., 2012. [http://idcdocserv.com/1414, viewedJuly 9, 2014.][6] Air Force Gentle Introduction to Software Engineering(GISE), www.stsc.hill.af.mil/resources/tech docs/gise.doc[7] Stephen Cass, S., Diakopoulos, N., and Romero, J.J.,The Top Programming Languages, IEEE Spectrum, July,2014.[8] x/[9] http://search.lockheedmartinjobs.com/[10] http://jobs-boeing.com/[11] http://jobs.raytheon.com/search/[12] 13] pportunities/broad-agency-announcements.aspx[14] http://www.arl.army.mil/www/default.cfm?page ctsheet.asp ? id tations/summary.do?method init&solId {7B28F838-C20E-7C9AE0E1-0F170B331AE9}&path open[17] www.abet.org[18] http://www.abet.org/eac-criteria-2014-2015/[19] 0] Brian R. Geiger, Joseph F. Horn,. Anthony M.DeLullo,. and Lyle N. Long, and Al F. Niessner, "OptimalPath Planning of UAVs Using Direct Collocation withNonlinear Programming," AIAA Guidance, Navigation, andControl Conference, Keystone, Colorado, Aug., 2006.7
8
For aerospace engineering programs, ABET [18] states: "Aeronautical engineering programs must prepare graduates to have a knowledge of aerodynamics, aerospace materials, structures, propulsion, flight mechanics, and stability and control." These are the "traditional" technology pillars [2] of aerospace engineering.
