ebook img

The Mathematical Intelligencer 29 1 PDF

60 Pages·2007·4.45 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview The Mathematical Intelligencer 29 1

Letters to the Editor The Mathematical Intelligencer encourages comments about the material in this issue. Letters tu the editor should be sent tu either of the editors-in-chief, Chandler Davis or Marjorie Senechal. Mathematics and Narrative Marjorie Senechal's article [1] is a delightful account of a meeting organised by the group Thales and Friends, attempting to explore the relationship hetween mathematics and narrative. What could he more beguiling than discussing this on the Aegean? But it seemed strange to read a paper on mathematics and narrative that doesn't mention the most commercially successful hooks ever written by a mathematician: Alice 's Aduentures in Wonderland and Through the Looking Glass. Neither is there any reference to the most successful paramathematical hook of recent years (and here I'm guessing even more wildly than in the previous sentence), Mark Haddon's The Curious Incident of the Dof!, in the Night-Time. Neither Carroll/Dodgson nor Haddon, who clearly know how to write something that people actually want to read, was trying to put mathematics across. They both produced beautifully written stories with a mathematical theme. Carroll just couldn't help being playful with mathematical and logical ideas: the mathematics in the background shines through. Haddon, who isn't a mathematician, has written a lovely story about a strange boy with some mathematical talent that (as it seemed to me) has captured a little of the feeling of what doing mathematics is like. Thales and Friends' website [2] is also a pleasure to browse through. I was particularly interested in the papers by Mazur [3] and Chaitin [4]. Mazur has made a heroic attempt to classify the different ways in which stories can he used in "mathematical exposition". Rut I think there is a problem here in his use of the word exposition. Stories and exposition don't seem to go together naturally: stories are surely more about exploration than exposition. Carroll is exploring logic in his Alice hooks: set up a crazy situation, apply the rules of logic, and see where we get to. And Haddon is explorinJ< the relationship hetween mathematics and autism, which, by the way. is exposed by James in [5]. Chaitin, in his paper, contrasts two views of mathematics: Hilbert's attempt to describe it as a closed, formal system of axioms, rules of deduction, and so on; and the Lakatos-Chaitin approach to mathematics as quasi-experimental. The Hilbert viewpoint demands exposition: here is mathematics all wrapped up in this formal system, now we must expose it. The Lakatos-Chaitin viewpoint suggests exploration: let's look around us, move off in an interesting direction, and see where it takes us. My suggestion is that, following Carroll and Haddon, you are much more likely to write a readable narrative if you can adopt the Lakatos-Chaitin-exploration point of view. The contrast between exposition and exploration, between a formal system and a quasi-experimental approach, seems very similar in spirit to the design/evolution dichotomy discussed in [6]. If you take a design point of view, then narrative, if it has any role at all , is merely a pedagogical device to sugar the pill or to set the mathematics in a wider context: a taxonomy for this is very well set out by Mazur. But with an evolutionary/exploration viewpoint, narrative is at the centre of the action. How could one provide any understanding of an evolutionary world better than by telling stories' No theory of economic development is going to give one a better idea of what happens in a market-place than the story of the evolution of the internet. And Haddon's story offers something that no formal theory of autism will ever give. Some businesspeople have adopted the idea of writing stories about the future to help them and their colleagues to understand possible developments in the business environment that may be of vital importance. These stories are called scenarios. (I would prefer to use the simple word story under all circumstances, but I have often encountered resistance unless I adopted more bureaucratic words like narrative, fiction, and scenario.) This approach was pioneered in the energy company Shell International (see [7]) in the early 1970s as a way of opening the minds of senior management to the possibility that the price of crude oil might one day rise above $2 per barrel , and has been used in Shell ever since. It has a number of advantages including: • Having more than one scenario reminds people that the future is unpredictable. Before reading Chaitin's © 2007 Springer SCience+ Bus1ness Medle, Inc., Volume 29, Number I, 2007 5 paper, I had thought this might not be relevant to mathematics, but now I'm not so sure. His idea of adding new axioms such as the Riemann Hypothesis (RH) in a quasi-experimental way seems rather analogous to the writing of different scenarios to explore the future. In fact I would like to challenge mathematicians to write two narratives about mathematics in one of which RH is true and in the other of which RH is false. If this turns out to be not possible, then I suppose the obvious next question is: does RH really matter? (I would also very much like to read a story set in a world in which the Continuum Hypothesis is false.) • The scenarios help to liberate people from "common sense" and from their prejudices. The different stories allow unusual ideas to be put forward and discussed as pieces of fiction rather than matters of life and death (which they can literally be, for example when this technique was used in a meeting between warring parties in South Africa towards the end of the apartheid era). This idea is reminiscent of the development of nonEuclidean geometries, a story that often inspires me when I write scenarios. • Occasionally, the same outcome crops up in the course of two quite different scenarios. This strengthens belief that this outcome might actually happen. Compare and contrast this with the famous result of Skewes and Littlewood that can be proved in two quite different ways depending on whether RH is true: far from having their belief in the result strengthened, some mathematicians, of course, have refused to accept that this constitutes a proof at all. The most significant benefit of scenarios, in my view, is in the understanding they bring to thinking about developments in the business environment, and hence an enormous improvement in the quality of discussions about unfolding events. (This is why [8] is called The Art of Strategic Conversation.) As events and trends happen, one can look at them and say, "Ah, yes, this is what one would expect in scenario A, but on the other hand that looks like scenario B". If you can say something like that, you have understood what is 6 THE MATHEMATICAL INTELLIGENCER going on. Chaitin says that understanding means compression, "the fact that you're putting just a few ideas in, and getting a lot more out". I think he's right in this, and that in practice the compression is often (always?) in the form of a narrative. When I say that I understand why somebody became angry, I mean that there is a narrative starring certain characters and featuring events and motivations, and that the anger fits into this narrative. When we say that we understand why the planets move round the sun in ellipses, we mean that this fits into the narrative of Newton, gravity, calculus, and so on. (This narrative will be more or less sophisticated for different people.) Here is an example from mathematics, which I think is archetypal. In [9], Singer says, "We should not be too surprised that mathematics has coherent systems applicable to physics. It remains to be seen whether there is an already developed system in mathematics that will describe the structure of string theory. [At present we do not even know what the symmetry group of string field theory is.]" Singer is trying to fit string field theory into a narrative. The narrative is called "Symmetry Groups", and Singer might think of this narrative as starring Galois and Einstein and a cast of thousands, or he might think of it, as with Mazur's story about rational points of elliptic curves, as a narrative of ideas. But it's a narrative. And if he finds out what the symmetry group of string field theory is, he will be justified in saying, "Now I understand!" in just the same way as a businessperson faced with the prospect of new environmental legislation can say, "Yes, I understand what's happening, it fits into one of my scenarios". If the symmetry group of string field theory is never found, then either string field theory will be abandoned or an entirely new narrative will need to be written. I can't stop without taking issue with a statement in [1]. (I think it is a remark made by a participant at the Thales meeting rather than necessarily the opinion of the author.) The statement is, "Popular math books must not mislead. They must tell the whole truth and nothing but the truth". If this were taken literally (and I imagine that advocates of the whole truth and nothing but the truth would like to be taken literally) it would simply mean the death of popular math- ematics. For the whole truth includes all the gory details, technical background, and arcane exceptions. This isn't popular mathematics, it's mathematics. Popular mathematics should certainly not mislead, but it can't afford to be cluttered up with the "whole truth". So what should popular mathematics do? It should be .faitf?ful to the narrative. The astute reader will have noticed that I have managed to write 1 ,500 words on mathematics and narrative without ever saying what I think a narrative is. But I'm in good company because as far as I can tell none of the Thales people has defined a narrative either. (Mazur does partly.) So on the principle of rushing in where angels fear to tread, here is a stab at a definition. "A narrative is a sequence over time of related specific events, emotions, or ideas designed to hold the attention of the reader, listener, or viewer. It is not the laying out of a general situation or theory in its entirety: but a good narrative will help people to gain a better understanding of the general situation. " Eric Grunwald Mathematical Capital 1 87 Sheen Lane London SW1 4 8LE UK e-mail: The Hidden Mathematics of the Mars Exploration Rover Mission UFFE THOMAS JANKVIST AND 8J0RN TOLDBOD 0 n January 4, 2004, Mars Exploration Rover (MER) A, named Spirit, entered the Martian atmosphere. The spacecraft, weighing 827 kg, was travelling with a speed of 19,300 km/h. During the next four minutes the velocity of the craft was reduced to 1 ,600 km/h at the meeting between the Martian atmosphere and the aeroshell of the craft. At this point a parachute was deployed and the velocity decreased to about 300 km/h. At a point 1 00 m above the Martian surface, the retrorockets were fired to slow the descent, and finally the giant airbags were inflated. The airhag-covered spacecraft hit the surface of Mars with a velocity of ahout 50 krn/h. The airbag ball bounced and rolled for about 1 km on the Martian surface just as Mars Pathfinder had done seven years earlier. When the landing module finally came to a stop its airbags were deflated and retracted and its petals were open. After six months in space the encapsulated rover, Spirit, could at last unfold its solar arrays. Three hours later Spirit transmitted its first image of the Gusev Crater to Earth. On January 15 Spirit left its landing module and drove out onto the surface of Mars. Ten days later, on January 25th, the entire scenario was repeated at Terra Meridiani with Mars Exploration Rover B, named Opportunity. 1 Introduction In March 2005 we spent a week at NASA's Jet Propulsion Laboratory QPL) as part of our joint master thesis 2 at the Figure I. Opportunitys heat shield and the shield's place of landing as seen from the rover. http: I /marsrovers. jpl. nasa.gov/gallery/press/opportunity/20041227a/ 1NN325EFF40CYLA3P0685L000Ml-crop-B330Rl_br.jpg mathematics department of Roskilde University, Denmark. The purpose of our stay was to conduct a small investigation of the mathematics in the Mars Exploration Rover (MER) mission being performed at JPL. Professor Emeritus Philip ]. Davis had more or less suggested such an investigation in an article published in 2004: Consider the recent flight to Mars that put a "laboratory vehicle" on that planet. [ . . . ] Now, from start to finish, the Mars shot would have been impossible without a tremendous underlay of mathematics built into chips 'This information in great part originates from http: I lnssdc. gsfc. nasa. gov ldatabaseiMasterCatalog?sc=2003-027 A and http: I lnssdcgs fc. nasa. govldatabaseiMasterCatalog?sc=2003-032A 2The thesis consists of the texts [5], [6], and [7] and can be found in its original Danish version as IMFUFA-text number imfufateksterlindex.htm 8 THE MATHEMAnCAL INTELLIGENCER © 2007 Spnnger SC1ence+Bus1ness Mecia, Inc 449 at http: 1 lrrunf. rue. dkl and software. It would defy the most knowledgeable historian of mathematics to discover and describe all the mathematics that was involved. The public is hardly aware of this; it is not written up in the newspapers. [2] Although we are not "the most knowledgeable historians of mathematics," we nevertheless decided to engage in Philip Davis's project. Unfortunately, newspapers are not the only place in which this wasn't written up. In fact, finding extensive literature on the mathematics of the mission was so difficult that we decided to base our investigation on interviews. Hence the long travel to Pasadena, California. While in the US we decided also to visit Davis at Brown University in Providence, Rhode Island, to discuss our pending investigation with him. Davis advised us to tly to gain an insight into the employees' personal motivations for working in the aerospace industry, as well as an understanding of the nature of the mathematical work performed at JPL [1]. We also decided to look at what might be referred to as the external influences on the daily work, such as deadlines and economic limitations-the basic work context. One of the more interesting aspects of our investigation quickly turned out to be the invisibility of the mathematics involved in the mission. The fact that the mathematics involved is hidden from the public may seem natural, hut parts of the mathematics of MER are also hidden from the scientists participating in the mission. In fact the hiding, or invisibility, of the mathematics in MER occurs on several levels, some intended and some not. The aim of this a 1ticle is to present some of the mathematical aspects of the MER mission and to discuss the way they are hidden in the mission, as well as the effect the work context had on the process5 Much of the account is built by letting the JPL scientists speak for themselves, i.e., by frequently quoting from our interviews. 1 JPL Scientists at Work We found that, in general, JPL's scientists arc people with the highest educational level who join the institution shortly after completing their university studies. They are driven by a desire to he part of the aerospace industry and a passion for planetary exploration. To some extent, they were also drawn to JPL by a fascination with the mathematical, physical, and engineering problems involved in space exploration; hut as a motivating factor this seemed secondary. Among the first to discuss the mathematical aspects of the work at JPL with us was Jacob Matijevic, a mathematician who had been with JPL for a long time. Particularly we discussed the modelling aspects of the work, which takes place before the actual mission is set in motion. A mission like MER is to a large extent about being able to predict how the technology onboard the craft is going to behave in space or in the Martian environment. Once the craft is flying it is impossible to make adjustments requiring more than a radio signal. Everything must therefore function as expected. Take for instance the Mars environment's influence on the instruments onboard Spirit and Opportunity. You have to have very precise knowledge about the distribution of heat inside the rover and how this affects the instruments. To acquire such knowledge, virtual models of the rovers are built in software so that the thermic conditions can be simulated. Such thermic models are typically based on a number of differential equations which are solved within the programs. The work for the JPL employee consists of building the virtual model of the rover. The exact method of solution which the program implements is secondary, as long as it works and is not too slow. According to Matijevic [ 1 1 ] you also need models of how the environment depends on the seasons on Mars to he able to predict the concrete influence on the instruments. These models are partly based on data from the different Mars orbiters and partly on concrete measurements performed on the Martian surface. The correctness of the surface measurements depends on how good the description of the instrument's behavior in the Martian environment is, and it cannot be guaranteed. By comparing the data from the orbiters with the surface measurements, a more accurate picture may arise; this may then be used to modify the models, so that they slowly become better and better. All of this is done in software. Regarding the models of how the seasons affect the Mars environment, it is probably fair to compare the work at JPL with that performed by an institute of meterology. Matijevic reported, When I first arrived here over twenty years ago there were still efforts to hand-implement certain mathematical models for certain applications. And there were specialist applications here for specialists in the applied mathematical sciences who worked here to make those 3An expanded Danish version of this article with a slightly different angle has also appeared in the Nordic mathematical journal Normal (13]. 4Transcripts of these interviews in full, along with our conversation with Ph1hp Davis, can be found in (7]. UFFE THOMAS JANKVIST was bom in BJ0RN TOLDBOD is a native of Roskilde, Den­ Copenhagen. He holds a master's degree from mark He holds a master's degree from Roskilde Roskilde University. He is now a doctoral can­ Universrty. He is now, as a conscientious objector didate there in the use of history of mathe­ to military service, doing his a�emative service matics in mathematics education. worl<ing at the computer department of the Royal Danish Library. Roskilde Universitetscenter Universitetsvej I Vesterbrogade I 19 A. 3.tv. 4000 Roskilde I 620 Copenhagen V Denmark Denmark e-mail: applications possible. But over time much of that has been incorporated in fairly standard and available simulation and modelling packages-computer packages. Expansions have been introduced slowly over time to these packages and that's basically how the engineers here do their job. Instead of going back to first principles they apply these tools . . . the foundation theories are from the eighteenth century to a large degree [ 1 1 ] . Hence a lot of work involving modelling and simulation is done at JPL, but all of it is done in software packages. This might lead one to suspect that JPL has its own staff of mathematicians developing such packages, but Matijevic informed us that the packages mostly come from commercial companies. A few days after our interview with Matijevic we had the opportunity to interview Dr. Miguel San Martin, an engineer with whom we discussed various aspects of the mission. He told with great enthusiasm about the challenges the scientists must overcome to make the rovers able to figure out their orientation on the Martian surface. San Martin explained that the navigation on the surface is based on a well-known technique which sailors have used for thousands of years on Earth-you look at the Sun. Together with a vector of gravitation, which can be measured by the rover, the position of the Sun in the sky provides the information necessary to determine the rover's position on the surface. San Martin didn't think of this problem as being very mathematical. In fact, he claimed that the majority of the mathematics involved in his work was very simple, and he concluded: The most important is that you have millions of these little, simple things. And that's the trick; to make them all work, and talk to each other and make sure that no parameter is tightened too much or too little. The complexity of the space problem is keeping it simple. [ 1 2] Present at this interview was Dr. William Folkner, a physicist and our main contact at JPL, and he elaborated on this view: Well, you hit a lot of mathematics in your descriptions, right, because you need to know the positions of the axes of Mars around the Sun as a function of time and you need to know what the orbits around the Sun and the Earth were. There is a lot of mathematics hidden in what you just said. [ . . . l We've worked all that out for us in the tables. So to know where the Sun is now, you just look it up. Somebody had to figure it out the first time. [ 1 2] Folkner's answer illustrates why the question of what mathematics is used in MER is difficult to answer. Knowledge that mathematics previously made accessible can over time become such an integral part of our conception of the world that we no longer connect it with mathematics. The trajectory of Mars is a good example of this: Is it mathematics to look up a table to see the Sun's position relative to Mars at a given time? Perhaps not, but the making of such a table is a mathematical problem. Thus the mathematics at JPL is often disguised as "common knowledge." MER Work Context Incredibly high reliability is demanded of the work performed at JPL. A single mistake in a piece of technology or an algorithm may have serious consequences and in the worst case may result in several years of wasted work for hundreds of people. All of the work being done at JPL is therefore subject to careful development and testing. We asked Jacob Matijevic about the development of the parachutes for the rovers, partly because we thought there would be little interaction between the parachutes and other devices. That is, we thought this would be a "simple" task. Matijevic, however, revealed more: We did drop tests. We did wind tunnel tests with the parachutes. But even before this time it was through models of the profiles of these devices that we came up with things like what the entry angles would he, what sorts of release points should we be looking at, as well as designing the algorithm that checks for height above the surface and finding out at which time to deploy the parachute and at which time to fire the rockets for slowing the descent. All of this was based on what we expected to be the environmental profile that the vehicle would see as it came down to the surface. So this was all done in simulation. [ 1 1] Figure 2. Guided tour at JPL. Left: Uffe and Dr. Albert Haldemann, who showed us some of the facilities. Right: Visiting JPL's museum for earlier space missions with Dr. William Falkner. 10 THE MATHEMATICAL INTELLIGENCER Figure 3. The guided tour takes us by JPL's "sandbox" where rovers are test driven. Left: Bjorn in front of the sandbox. Right: A replica of a Mars Exploration Rover used for test drives at JPL. Reliability is paramount for any mission. If the choice stands between two different approaches to a problem, a space scientist will be inclined to choose a well-known, well-tested solution over a new-perhaps more efficientsolution which has not been thoroughly tested in the context of the mission. Our interview with Dr. Jon Hamkins, one of JPL's leading coding theoreticians, also confirmed this for the error-correcting codes used in MER [4]. Mathematics and technology which had been onboard an earlier mission is considered to be safer and therefore makes a more attractive choice. This approach is taken in all aspects of the missions. Of course some development takes place from mission to mission hut only at a pace that makes extensive testing possible. Jacob Matijevic called this ··steady progress'' [ 1 1 ] . New ideas introduced into the missions will he at least '5-10 years old at launch time, because they must he laid down when the missions are first planned. In the Figure 4. case of error-correcting codes, a lot of the mathematics involved has to he implemented in hardware for speed. Generally hardware is much more expensive to replace than software, so the gain of introducing a new error-correcting code has to he considerable in order to balance the expense of the substitution . The grand scale of a project like MER also means that the work performed by different departments must be completed at specific deadlines. Not surprisingly, deadlines may serve as a stop block for the development of new ideas. You cannot promise to use a new method if you are not sure that there will be enough time to test it. There is another factor working against the introduction of new methods: In recent years .JPL has gone from a small number of large and expensive missions to a large number of small hut less expensive missions. For instance, the Pathfinder mission of 1996 had a total budget of 265 mil- Left: Wind tunnel test of the MER landing module parachute. http: //www.nasa.gov/centers/ames/images/ Right: MER landing rnoduit: airhags. http: //photojournal. jpl.nasa. gov/jpegMod/ content/79641main_picture_2. jpg PIA04999_modest. jpg © 2007 Spnnger Sc19nce+Bus1ness Media, Inc. Volume 29, Number 1. 2007 11 Figure 5. Guided tour at JPL. Left: A JPL photo of a MER rover testing prior to launch. Right: One of JPL's laboratories. lion dollars 5 whereas the Viking missions of the seventies had a budget of around 8 billion dollars. The MER mission was more expensive than Pathfinder, hut still nowhere near the Viking budget. This means that if anything from a previous mission can be used again, there are huge amounts of money and time to be saved. Many of the cheaper missions must necessarily rely on reuse from earlier missions. The nature of the space missions in general was summed up in the following way by Falkner: Everything is a cost-benefit analysis. The whole space system is a cost-benefit analysis. [3] To this point we have mostly focused on how external factors influence the mathematics of a space mission. We now turn to two examples of concrete mathematical problems in the MER mission. Still the purpose of our selection will be to illustrate common features of the mathematics in a space mission. coded, then it goes through a block interleaver and then it's convolutionally encoded [4]. Thus MER's coding system consisted of two combined, or concatenated, error-correcting codes; a Reed-Solomon code and a convolutional code. Reed-Solomon codes are so-called algebraic codes, whose code symbols come from a Galois field. ReedSolomon codes are linear and cyclic. If the data words to be encoded are of length k and code words are of length n, an (n,k)-cyclic code, with code symbols from the Galois field F q• is a cyclic subspace of the vector space Fq. Reed­ Two Selected Mathematical Problems of MER Solomon codes can either be defined as special kinds of cyclic codes with specific generator polynomials or by means of Fourier transforms. On the one hand, convolutional codes are not as mathematically well understood as the Reed-Solomon codes and not as easily defined either. They can, however, be defined by the use of formal Laurent series. On the other hand, convolutional codes are very efficient and therefore very often used in technology. They are excellent for correcting single-bit errors, the kinds of errors which most often occur from interference in deep space. Unfortunately the decoding of convolutional codes 6 often results in a run of consecutive errors, so-called burst errors. Fortunately Reed-Solomon codes excel exactly in correcting burst errors, hence the concatenated system. The reason for first encoding the data with the Reed-Solomon code and then with the convolutional code is that the decoding procedure must be the reverse of the encoding procedure. Block interleaving is a technique used to ensure that the hurst errors from the convolutional decoding are no more severe than what the Reed-Solomon decoder can handle. One of the main purposes of the MER mission was to take photographs of Mars. Before these photos were transmitted to Earth they had to he compressed. The imagecompression technique primarily used in MER is called The /Ct-1? Progressive Wavelet Image Compression, in short just ICER, and was developed at JPL by Drs. Aaron Kiely and The mathematical emphasis of our investigation was the two coherent mathematical theories called channel coding and source coding, which deal with reliable communication and compression of data (including images), respectively. Formal definitions of the specific codes involved would take us beyond the scope of this article. We therefore merely indicate the areas of mathematics involved. The signals transmitted to and from Mars are subject to interference during their travel through deep space. Such interference of a binary signal may result in bits becoming altered. The communication between Earth and the rovers needs to be reliable. The problem of interference is solved by way of channel codes, which make it possible to correct altered bits in a message; hence the codes are also called error-correcting codes. The system used in MER depends on two different codes used in combination. Jon Hamkins explained: The majority of the missions flying now are concatenated. So the data comes in and is Reed-Solomon en- 5http://nssdc.gsfc.nasa.gov/database/MasterCatalog?sc�1996-068A 6JPL 12 uses the so-called Viterbi algorithm in its convolutional decoder. THE MATHEMATICAL INTELLIGENCER Matthew Klimesh. The word "progressive" refers to progressive fidelity compression. In such compression a lowquality approximation of the photo is first transmitted . Afterwards hits are transmitted in such a way that the quality of the photo is gradually improved. When all hits are transmitted the reconstructed image equals the original image. By stopping the transmission before it is complete, lossy compression (i.e., compression with loss of data) can be obtained. In this way ICF.R supports lossless as well as lossy compression, even though it was entirely used for lossy compression. For lossless compression MER relied on the commercial compression algorithm LOCO. The ICER algorithm, like many other image-compression techniques. overall consists of three stages: preprocessing, modelling, and entropy encoding of data. Kiely explained: We got data coming in, an image or whatever it is, and then some sort of preprocessing stage, for example a wavelet transform plus quantization or a discrete cosine transform or something. The goal is that it doesn't perform any compression, and in fact it is often a lossy process, it might throw out some of the data. but the idea is to process the data in a way that makes it more receptive to compression through the entropy encoder. The entropy encoder is sort of the engine. Given some sort of probabilistic model of the source, it compresses data or represents it in a more efficient way through something like a variable-length code. That is sort of the big picture of what is going on . So for example for ICER what is going on is mostly a probabilistic transform. For LOCO it is in essence trying to project a probability distribution on the next pixel that it is about to encode based on what it has seen in the nearby neighbors. (9] ICER uses a wavelet transform that closely resembles a Haar tran.iform, a context model also known as a Markov model, and the majority of the entropy codes used by the entropy encoder are the Golomh codes [R]. The LOCO algorithm is a hit different from ICER in that it does not have a preprocessing stage which is typical for losslcss compression techniques. It does use context modelling, and it uses both Golomb codes and Huffman codes [14] . The Haar transform i s an invertible transform which makes hoth lossy and lossless compression possible. An image with n X n pixels is naturally represented by a n X n matrix M containing integer values. By the invertible transform, say T: R" � R", which is first performed on rows and then on columns of M, a shift of basis from the euclidian basis to another orthonormal basis is made. The elements of this basis are called wauelets. The transform splits the image into a lowfrequency suhimage and a high-frequency subimage, concentrating the majority of the energy in the low-frequency area, thus making efficient lossy compression possible. In a context or Markov model the likelihood of a symbol's being encoded depends on the previously encoded symbols; i . e . , the model is said to have a memory. This is a very common situation in digital images, where the value of a certain pixel may indeed depend on the values of the surrounding pixels. ICER's context model maintains a statistical model with the purpose of estimating the likelihood of the next hit's being a zero bit. Huffman codes are variable-length prefix codes, which means that they can assign shorter codewords to more frequent data symbols and that no codeword is a prefix of another codeword. Golomb codes make u p another family of codes which are parametrized by an integer m > 0 and therefore are often written as <§m· A Golomb code encodes integers under the assumption that the larger the integer the less likely its occurrence is. It can be shown that Golomb codes make optimal encodings for geometric probability distributions of non-negative integers, which make them attractive in many different contexts. The above presentation of mathematics in MER is, of course, merely scratching the surface. Other mathematical theories that we came across during our interviews at JPL include the "Lost in Space" problem, pinpointing the craft's position in space Jt a given time; Kalman filtering, estimating incomplete and interfered-with data from the craft's different sensors; the Hohmann trajectory, the trajectOiy hetween Earth and Mars which calls for the least amount of energy at launch; and of course control theory. Discovering every little piece of mathematics put to use in the MER mission probably is an impossible task, as Folkner im!icated: There is mathematics in everything. There is control theory, aerodynamics, orbital dynamics, Newtonian gravity , bodies going around the Sun. We use general relativity, that's mathematics of physics. [ . . . ] Linear algebra is a field of mathematics we use all the time. Matrices. That's in the control theory all the time. There is Riemannian geometry in the general relativity. Calculus. [ 1 2] Instead, we shall now turn to the theme of this article, the hidden mathematics of MER. Hidden Mathematics of MER Figure 6. Spirit's landing module as seen from the rover. http://www.seds. org/-spider/spider/Mars/Hi-res/ mer a lander. jpg When you view a space mission from the outside, it is clear that the mathematics involved is to a certain degree hidden (or invisible). We had imagined that once we entered the JPL facilities the mathematics of MER would suddenly become more visible. Our investigations showed otherwise. © 2007 Spnnger Science+Bus1ness Media, Inc , Volume 29, Number 1. 2007 1 3 The mathematics of a mission like MER is hidden on several different levels. To most people reading about space missions in a newspaper, the mathematics is hidden in a lot of technology whose mode of operation is seldom discussed. Partly, the media fear boring their readers; but even when they try, it is difficult to communicate abstract mathematics to the uninitiated. This type of invisibility seems obvious. More interesting, the mathematics to some extent also is hidden inside the walls of JPL. As we pointed out, a large part of the mathematics for solving the mission's problems is embedded in software packages-and therefore hidden. The packages are often commercial packages, developed outside JPL. To users of such packages it is important to know how to use the packages correctly, and this most certainly requires mathematical knowledge, but it is not necessary for a user to understand the specific mathematics in detail. One might say that the mathematics is outsourced, and that the implementation in software packages contributes to hiding the mathematics from the people at the ]PL. As the conversation between Falkner and San Martin reported above illustrates, a consequence of the mathematics being hidden is that to some degree it is being ignored. The virtual m o dels of the rovers and the models of the Martian environment of which Matijevic speak above rely also on the mathematical technique of modelling. If the scientists of JPL who prepare these simulations do not regard the modelling as mathematics-and our investigation suggests that to some degree they may not-then this mathematics will also be hidden from them. Anyway, the JPL projects are so vast that a distinction between being inside or outside is somewhat meaningless. A scientist who works with a specific part of the mission of course has an overall perception of the mathematics involved in other parts of the mission, but as the following quote from one of our interviews with Falkner shows, the details are sometimes quite hidden. Falkner's first remark was a joke intended to illustrate that the use of mathematics in a mission like MER is so extensive that it would be easier to expose the mathematics not involved. We don't do any map theory for instance. Four-color map problem. You know that? We don't do any of that. We don't really use any abstract algebra, group theory, and that sort. Except in the channel coding. They use abstract algebra and group theory in that? The Reed-Solomon codes are based on Galois Fields. That's news to me. I didn't know that; all right, interesting. [ 1 2) Worth noticing is that Falkner says, " They use abstract algebra and group theory in that?" This suggests that the different departments of JPL may not have a large degree of interaction. In fact William Falkner and Miguel San Martin had never met each other before we interviewed them together. The point here is that the vastness of the project contributes to hiding the mathematics from the employees. One last kind of invisibility which must be mentioned is that certain areas of the missions are classified: some things are deliberately hidden. Our experience at JPL was that the institution generally was very open to giving in- 14 THE MATHEMATICAL INTELLIGENCER formation and answering questions, but also that security, control, and classification were part of a normal work day. Thus when we talked to Mark Maimone about the amount of control theory used for steering the rovers, he replied at one point, "I don't know what has been published about the details of those algorithms, so I can't tell you anything about that" [ 10). We have not examined the motives for such classification any further, but it seems fair to say that classification makes it harder to discover the mathematics involved and therefore contributes to the hiding of it. When discussing the hidden mathematics of a project like MER, one question springs to mind: Would the applied scientists of MER have been able to benefit from knowing more about the mathematical tools they use in their work? Perhaps. Our investigation does not imply a unique answer to this question. If we restrict our attention to the modelling and computer simulation aspects of MER, there might be something to say on this topic. Whenever working with commercial software and standard packages users may have only a general idea of the (mathematical) elements involved. In a large-scale project like MER, different people, and different departments work on different aspects of a computer simulation. Indeed, scientists do not always know in detail what is going on 'on the other side of the fence. ' In recent years the engineering industry has been experiencing a 'paradigm shift' in the way engineers work: the testing of scale models has now been replaced by computer simulations, and the engineer most likely does not know anything about the model(s) on which the simulation is based. Now, this is a scary scenario, for wrong applications or wrong interpretations of the simulations may lead to disastrous mistakes. In such situations the applied scientists would most definitely benefit from knowing more of the ramifications of the (hidden) mathematical tools they use. That they don't is quite paradoxical: the space missions are devoted to minimizing mistakes, yet the nature of the space missions may itself inflict errors-just because the mathematics in use is hidden. Conclusions A lot of the mathematics of MER is hidden and not only from the public but even from the applied scientists working on the mission. As briefly sketched above, for the scientists, this could be disastrous in a worst-case scenario. The hiding of mathematics, both in our everyday life and within science itself, is a matter not often discussed in public-which in itself is a disaster, taking into account the consequences the hiding of mathematics might have for the public. We like to think that this article may help let in some light. Another question raised by our work is that of beliefs in mathematics. Only occasionally are the beliefs of mathematicians discussed. We found repeatedly that mathematical elements of MER are not actually considered to be mathematics among the applied scientists themselves, not on first hand anyway. Is this due to the fundamentally different views of what mathematics is between applied scientists (including engineers) and pure scientists of the 20th century? We do not know. Finally, we comment on the nature of the mathematics involved in MER. Because of the extreme nature of a Mars mission, one might expect "extreme'' mathematics, mathematics developed for the sole purpose of this mission. This does not seem to he the case. We did not come across any basic innovations in mathematics as a result of the MER mission. The MER mission is based on well-established mathematical theories and disciplines. Some of these date back to the 18th century, but a lot of them arc also from the 20th century: the Hohmann trajectory is from 192'i, the convolutional error-correcting codes used arc from 19'i4, and the Reed-Solomon codes are from 1 960, the Lost in Space problem and Kalman filtering arc also from the 1 960s. Image compression relies on the Haar transform from 1 910, Huffman codes from 19'i2, Golomb codes from 196(J, and wavelet theory from the 1990s. Now, this is not to say that basic research in mathematics is not used. Rather it suggests that technological and mathematical developments seem to he on different timelincs so that a stop in research in pure mathematics would damage technological development. And since so many of the mathematical theories in MER are from the 19(JOs, one might think, that the timelines in the case of the space industry arc no more than fifty years apart. Feedback also goes the other way: basic research in mathematics can he inspired by problems in applied sciences and technology [lJ-therefore also from missions like MER and institutions like JPL. But the major role of JPL seems to he that of consumer of already developed applied mathematics. From start to finish MER is an example of the application of mathematics. But not just any example. It is an extraordinary example of what can he accomplished with the mathematics at our disposal. BooiS-Bavnhek for suggestions and comments on the original thesis. BIBLIOGRAPHY [ 1 ] P. J. Davis. Interview with Professor Emeritus Philip J . Davis, March 6th, 2005. Recorded at Brown University, Providence, R l . [2] P. J. Davis. A Letter t o Christina of Denmark. EMS, pages 21-24, March 2004. [3] W. Falkner. Interview with Doctor William Falkner, March 1 7th, 2005. Recorded at JPL, Pasadena. CA. [4] J. Hamkins. Interview with Doctor Jon Hamkins, March 1 4th, 2005. Recorded at JPL, Pasadena, CA. [5] U. T. Jankvist and B. Toldbod. Mathematikken bag Mars-missio­ nen - En em pi risk unders0gelse af matematikken i MER med fokus pa kildekodning og kanalkodning. Master's thesis, Roskilde Uni­ versity, October 2005. Tekster fra IMFUFA, nr. 449a. [6] U. T. Jankvist and B. Toldbod. Matematikken bag Mars-missio­ nen - lndfmelse i den grundlceggende teori for kildekodning og kanalkodning i MER. Master's thesis, Roskilde University, October 2005. Tekster fra IMFUFA, nr. 449b. [7] U. T. Jankvist and B. Toldbod. Matematikken bag Mars-missio­ nen -Transskriberede interviews fra DTU, Brown University, MIT og JPL. Master's thesis, Roskilde University, October 2005. Tek­ ster fra IMFUFA, nr. 449c. [8] A. Kiely and M. Klimesh. The ICER Progressive Wavelet Image Compressor. lPN Progress Report, 42(1 55) : 1 -46, November 2003. [9] A. Kiely. Interview with Doctor Aaron Kiely, March 1 4th, 2005. Recorded at JPL, Pasadena, CA. [1 0] M. W. Maimone. Interview with Doctor Mark W. Maimone, March 1 7th, 2005. Recorded at JPL, Pasadena, CA. [1 1 ] J. Matijevic. Interview with Doctor Jacob Matijevic, March 1 4th, 2005. Recorded at JPL, Pasadena, CA. ACKNOWLEDGMENTS First of all we would like to thank all the employees at JPL who set aside their duties to talk to us, especially Bill Folkner. who arranged most of our meetings. We also thank Phil Davis for taking such an extraordinary interest in our investigation and for commenting on this article. Furthermore we thank Tinne Hoff Kjeldsen, Man-Keung Siu, and Chandler Davis for helpful suggestions about how our investigation should he presented here. Thanks also to Anders Madsen and Bernheim [1 2] M. San Martin and W. Falkner. Interview with Doctor Miguel San Martin & Doctor William Falkner, March 1 4th, 2005. Recorded at JPL, Pasadena, CA. [1 3] B. Toldbod and U. T. Jankvist. Reportage fra en Mars-mission. Normat, 54(3): 1 0 1 -1 1 5 and 1 44, 2006. [1 4] M. J. Weinberger, G. Seroussi, and G. Sapiro. LOCO-I: A Low Complexity, Context-Based , Lossless Image Compression Algo­ rithm. Proceedings of the IEEE Data Compression Conference, pages 1 -1 0, March-April 1 996. "Celebrated Turtle" Plea�ed and proud as \'d.' were to present the mono-monostatic and P. L. Ya rkon ·i, a nd gratefu l as we are to the artbt for ih we '' nml • ( ,-ol. 2H, no. "!. 3-1-3R) of Gabor Domokos beautiful pr •s ·ntation on the cover of th ' issue. cr •n't quite prt'pared for the media frenzy it excited in H ungary and far herond. Th ·re ar • dozens of ar­ tid ·s ,1hout it-sec. for example, http; WW\\ .lahoodlc.hu nc news ne,,-s_archi,-e singlc_page article 1 1 hungarian_ sc ?cHash� i2Sf3fct7b The story has h L'n pick •d u p h} Reuters and broadcast around the world: see. for e. ample, http:/ WW\\ . msnlx.msn.com id 1 - 1 20210/. And go to Imp:/ ww\v .sciam .com and d ick on Biolog} . Gabor comments. "I am tell i ng vou this to sho\\' hem much impa t your \York ha., in u nli ke! corners of the world." \\'e r •ply . with all due t hanks, that the \\'ork '' .1s the authors . not our. , and that Hu ngary, if indeed it is a corner of the world, is a rel::nin:l} likely corner. -The Editors © 2007 Spnnger Sc1ence+Bus1ness Media, Inc , Volume 29, Number 1 , 2007 15

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.