Robohub: What funding scheme is most conducive to creating a robotics industry?

money robotRobohub just posted a great series on optimal funding schemes for robotics start-ups.  I highly recommend reading it.  I believe that it probably represents the best collective wisdom in our industry.  Frank Tobe probably has the most informative response for someone actually looking to raise money: robotics is still at the point where you need to appeal to individual personalities who see it and get it, or find a government customer.  However, I thought that all the authors raised thought provoking points.  Here are the follow-up questions that I posed:

Rafello D’Andrea:  What structural and cultural changes need to be made to robotics departments so that they become as entrepreneurial as computer science or biochemistry departments?

My own observation is that here at CMU–one of the most prolific robotics start-up hot beds–that robotics is pretty theoretical and academic compared with other engineer disciplines, particularly other disciplines in the computer science school.  The revolving door between industry and academia just doesn’t happen in robotics the way it does in other disciplines.  How do we get industry thinking into robotics departments?  After staying close to the university for 40 years, it is going to be hard to change the culture of the robotics departments, however I think that universities that succeed have a chance to maintain or overtake the currently established leaders in the field.

Henrik Christensen:  If much of the benefits from robotics R&D accrues to parties who didn’t do the research—whether competitors or society at large, economics tells us that subsidies are not only appropriate, but necessary, to get to the socially optimal level of investment. What portion of the gains from commercial robotics R&D is controlled by the company that does the research?  How does this compare with other industries?

I know the Georgia political climate is such that private industry is always the answer.  We all agree on the need for more private investment, but if robotics companies have trouble capturing the value that they create, we need to do one of two things:  1) Either subsidize their research in some rational way that creates the most social gain or 2) adjust intellectual property laws so that more of the benefit of robotics R&D accrues to companies making the investment.  Some econometric research is probably in order here… any econ Ph.D. candidates reading?

Mark Tilden:  Doesn’t your suggestion of investing in crowd funded start-ups point at the opposite of needing more innovative roboticists?  If crowd funding is the shining star in our industry, wouldn’t that suggest that our roboticists are plenty innovative—as high end research is not required to make marketable stuff—but rather our entrepreneurs and business managers are behind the power curve?

Obviously, market traction is the key.  Financing is for companies is in some way just a loan to future consumers–even if the consumers don’t know it.  This question of what’s the real roadblock to creating more successful robotics start-ups is a key one.  I’ve made my belief that the robotics “parts bin” has plenty of technology in it pretty clear on this blog as well as my belief that robotics has a shortage of qualified entrepreneurs and managers.  The problem is not on the engineering side, it is with those giving directions to engineering.

Frank Tobe:  If the individual / angels / VC route is more of the direction that we want robotics to go in, what do the special people that you point to in your response see that other investors don’t see?  Or are they doing something different?   What is the barrier to other investors who might want to do the same?

If robotics is at the point where it is being funded by visionaries, how does one go about finding, cultivating, or creating more?  Are the visionaries right or is their compass off?  I don’t have good answers to this, but I do think that robotics seems to require a more comprehensive understanding of engineering, current business practices, and what the future should be than most other industries do.  That said, one would expect that there are extraordinary rewards for solving these hard problems, unless some of the basic economic problems that I want to suggest in my question to Henrik Christensen exist.

Nicola Tomatis:  Software and biotech companies aren’t cheap to build in absolute sense either, but they are called capital efficient by investors.  Financially, robotics is probably more like software and biotech than it is like retail or [green] energy businesses—which really require a lot of money.   Is there data that supports the position that robotics is expensive compared to other capital efficient industries?

The part of this blog I’m most proud of is gathering the evidence to show, to a practitioner’s standard, that robotics companies are as  capital efficient as software companies, conditional on success for both.   While plenty of robotics companies waste investors money, I’m not sure that this that different from any other IP intensive industry.  However, whenever a software company fails we blame management or the market–but when a robotics company fails we blame the underlying technology.  We need to stop that.  It makes it harder for the next guy to start a robotics company–the underlying technology is there–we just haven’t made many companies with it yet.

It is not a secret any more…

This is what I’ve been working on that’s been keeping me away from the blog:

 

TerrAvion-Logo-textverticalpng

The Knowledge Economy Cash Anomaly, Part 3: The exciting conclusion

This is part 3 in a series.  Here are Part 1 and Part 2.

Tax Shields 

The organization of the knowledge economy is inclined towards creating great tax advantages.  Both start-ups and mature companies enjoy huge advantages that the resource economy does not enjoy.  Most investments can be expensed.  The companies grow fast enough that they create huge tax losses, even as they create extraordinary value for the owners.  Once they become mature global companies, their assets can be transferred almost costlessly to whatever jurisdiction offers the most favorable treatment.  Transfer pricing makes it almost impossible for authorities to tell where value was added.  Money generated off-shore can stay off-shore tax free indefinitely.  In contrast, resource economy companies have easily traceable assets, some of which require particular locations and may be quite literally fixed to that location.  Their assets are comparatively easy to tax, whatever form their assets take.

If this is the case, it follows that knowledge economy companies have huge tax shields from their operations.  To have these tax shields add value to the business, the CFO of a company needs a business that is low risk, earns the cost of capital after tax, and does not consume much management attention.  Investing in marketable securities seems like just the ticket.  The gain on securities allows the owners of the company to take advantage of the tax shields that would otherwise go unused.

Here is what we’ve been looking for all along.  A reason why cash is better off in the pocket of the company than in the pocket of the owner.  In addition, all the other reasons why a firm might hold so much cash are still active and valid.  Full use of tax shields would be a driving factor for keeping cash on the balance sheet.  The discount rate for tax shields is low and even if only gets used every few years, it adds to the wealth of the shareholders.  For a founder, cash on the balance sheet capitalizes an otherwise unused tax shield, provides diversification, defends the core business, and enhances the value of R&D investments by its mere presence.

The question for further study would be when we would expect to see these benefits diminish?  Can we empirically test which of these hypotheses are most important in guiding payout policy?

The Knowledge Economy Cash Anomaly, Part 2

This is a continuation of Part 1.

Option Value of Cash on the Balance Sheet

This theory of the cash anomaly posits that the returns from R&D are high, but also highly uncertain.  Every once and awhile, the R&D of a company will produce a really high value project that requires massive investment and possibly acquisitions to use in combination with the asset.  The problem with R&D as an economic asset is that it is very difficult to sell or even be exploited by organizations other than the organization that developed it.  Unlike discovering oil, it is not clear even after discovery of a project that another firm could develop the project to create economic returns.

Because exploitation relies on unique capabilities inside the firm that are only poorly understood outside the firm, their economic value is harder to forecast.  This violates the costless symmetric  information condition of efficient markets is violated, unlike the projects of old economy companies, where the market has a reasonable expectation that it will understand the value of the project.  This uncertainty introduces huge frictions if projects need to raise new capital. Therefore, if a company has R&D projects, the value of that project stream is greatly enhanced if the company also has a means of financing the projects that does not require subjecting those projects to the friction of market financing.  These frictions are both directly financial in the form of more returns to new investors and intermediaries, and also temporal.  In winner takes all markets, which many technology markets are, temporal costs are huge.

The option value of cash on the balance sheet could be huge, however, we would expect more tech companies to at least on occasion, expend all their cash and perhaps even borrowing capacity when they exercised options if this were the case.  This is common in growing technology companies.  Mature tech companies, rarely, if ever come close to expending their investment capacity.

I’m skeptical of this explanation.  Why does Google need to hold enough cash to buy Yahoo or Facebook in cash, if they are never exercise the option to do so?  When was the last time you heard that a company was undertaking a project with more than a billion dollars of expenditures in year one of the project?  These kinds of companies can make acquisitions with stock, invest over time out of future cash flows, and they even have relatively low cost borrowing capacity should it be required.

Cash Poor at Home

Recently, much has been made of the U.S. companies that are parking cash overseas to avoid the tax when they repatriate it.  Many companies are cash poor in their U.S. entity, but their consolidated balance sheet shows a lot of cash.  This cash can’t be repatriated for distribution without a large tax bill.  This is the worst of all possible worlds from a policy perspective, but it doesn’t seem to afflict tech companies as much as industrial conglomerates.

(BTW, Congress doesn’t need to capitulate to corporate demands for no tax on foreign earnings.  All it has to do charge the companies income tax on their cost of capital for any overseas investments, then true up when the companies bring cash home.  Particularly if the law slightly over estimated the cost of capital, or ignored the cost of capital on financial assets in the WACC calculation, so that repatriating funds usually triggered a small refund rather than a small bill, you could just sit back and relax and watch them all bring their cash home while still paying tax.)

Distress Costs

The final explanation I’ve heard offered is the idea that since most of the investments of a technology company are in workforce and R&D, the costs of financial distress are huge.  Not only that, but the costs of financial distress can manifest themselves long before bankruptcy is close.  If managers are cutting benefits or tightening R&D activites, and the costs are not properly captured by accounting frameworks.  New talent goes elsewhere, the best old talent leaves, R&D becomes less creative, less real economic capital employed stealthily decreases without the accountants noticing.  However, CFOs are smart, they know this–even if the accountants don’t.  They keep cash on the balance sheet, employee benefits generous, and 10% time meaningful.  This prevents the stealthy erosion of the real assets of the company, by the prospect of distress, which the intelligent and savvy workforce is acutely aware of even if they don’t conduct formal analysis.

But there is one more reason…

In part 3, I will outline how holding cash creates economic value, regardless of and in addition to, all these explanations.  Go to Part 3.

Cognex [NASDAQ:CGNX]: Economic Valuation of A Robotics Company

I prepared this valuation for Prof. Joel Stern.

If you would like to see a chart or table with a white background, click through it twice.  Use the back button to return to the article.

Executive of Summary

Cognex is correctly valued in the market.

A aiagram from a machine vision patent assigned to Cognex

A diagram from a machine vision patent assigned to Cognex

Overview of Cognex

Cognex is a machine vision systems corporation—they focus on computers which can see—particularly in industrial automation applications.  Originally an MIT spin-out, whose name stood for Cognition Experts, they are headquartered in Natick, Massachusetts—though one of their two main divisions is in the Bay Area like a respectable technology company should be.  They have been public since 1989 and have been paying an extremely modest dividend since 2003.

CGNX Share Price Chart

Figure 1 – Source:  Google Finance

As of close on December 7th, Cognex stock was trading at $36.62 a share with 42,961,000 shares outstanding and a market capitalization of $1.573 billion.   Their revenues are well diversified with 66% coming from outside the United States and the top five customers only account for 7% of revenue.  Like most robotics companies, Cognex has no debt and exhibits the cash anomaly of the knowledge economy.  For tax reasons, Cognex is planning to pay a large 4th quarter dividend, but before paying the dividend, Cognex will have over $400 million in cash and securities on its balance sheet.  Cognex’s non-financial, GAAP capital, net of operating liabilities was only about $200 million and of that $80 million was goodwill.  Contrary to popular wisdom, it does not take a lot capital to build robots.

Cognex is a classic, mid-sized, public robotics company if there if ever was one.  Financially, it looks very similar to other successful robotics companies like Brooks Automation (BRKS), iRobot (IRBT), Aerovironment (AVAV), and to a lesser extent Intuitive Surgical (ISRG)—although none of these companies are direct competitors.

Cognex has unique technologies, a portfolio of successful and related products, and a habit of expanding its business with both organic growth and prudent, related acquisitions.  The macroeconomic trends of the coming decades probably favor Cognex.  The growth of on-shoring, higher labor and environmental standards, rising third-world wages, continued growth of the global middle class, and the increased pace and automation of supply chains all favor the growth of Cognex’s business.  There is some threat of emerging competition or economic disruption from start-up companies like ReThink Robotics, but Cognex’s cash and industry relationships make it equally likely that they are the distribution and exit strategy for such start-ups.

Valuation Process

The valuation process relies on data gathered from market reporting and the SEC’s EDGAR database.  Historical returns allowed me to compute the cost of capital.  Following this, I made adjustments to discover Cognex’s historical assets and economic returns to assets.  I assumed that the 7 year historical return, approximately one economic cycle, would be a good guide to future returns as this is not Cognex’s first economic cycle.  This means that we are assuming that Cognex returns 21.3% on its economic assets every year.

I used a somewhat roundabout way to get investment.  First, I assumed that the GAAP assets required to produce these sales would remain unchanged and so depreciation would exactly equal GAAP investment.  Compared to other robotics and tech companies Cognex has too many GAAP assets, see Figure 2.  To estimate future R&D spending, I observed Cognex has been remarkably consistent in spending 14% of gross revenue on R&D, so I backed into gross revenue from the economic return on assets by assuming a fixed ratio from historical data.  From there, I took 14% of gross revenue and added it to capitalized R&D.  From this capitalized R&D figure, I removed 1/12 annually for obsolescence, to arrive at a capitalized R&D figure.  This figure was added to GAAP non-financial assets to get the economic assets of the firm.

Reader, my apology for overuse of this chart

Reader, my apology for overuse of this chart

Figure 2 – Source: 2011/2012 10-Ks on EDGAR as of July 2012

From this forecast of the company growth, I used three valuation methods.  First, I estimated a free cash flow, which is the economic return of the assets of the company less the addition to capitalized R&D.  Because they have no debt and no GAAP investment beyond depreciation, this is equal to Cognex’s operating profit.  Next, I calculated the economic value added, this is the spread on the total economic assets employed by the company in any given year.  I calculated both of these methods for the next 20 years, with a perpetuity value beyond the forecast period.  Finally, I calculated a long form economic value driver model of the firm.  For this, I ran the calculation two ways.  One way, the forecast period is 20 years, the other has an investment period of 10 years.  The ten year period brought the value in line with the other methods.  This may be a consequence of the way that I dealt with the changing investment amounts.  However, the long form is mostly intended to talk about the sources of value in the stock price, not accurately predict what the price should be.

Cost of Capital

To estimate the cost of capital for Cognex, I regressed the monthly returns to Cognex over the ten year treasury return for the last five years against the equity premium of the Russell 3000.  The result is below in table 1.  The alpha is not significant—and even if it was, this alpha could not be expected to be permanent—however forcing it to zero does not yield a significantly different beta, so I used a beta of ~1.38.

Cost of Capital Regression

Table 1 – Regression of Cognex Premium Returns to Russel 3000 Premium Returns

This beta times a future equity risk premium of 6% and on top of a ten year risk free rate of 1.626% results in cost of capital 9.89%.  Since Cognex has no debt, this is the weighted average cost of capital as well.  The ten year bond may not be a perfectly appropriate choice given our forecast period of twenty years, but it should be an adequate estimator for our purposes.  Using the 30-year yield would raise the cost of capital by about 1% to be almost 11% instead of just under 10%.  Given the economic spread that Cognex returns, this would change the valuation by about 10-15%, but it probably wouldn’t change many of the company’s investment decisions.

Free Cash Flow Valuation

Using the method above, I prepared a forecast of the free cash flows Cognex can be expected to produce.   The table below shows the forecast with the intervening years truncated.  Of course this forecast does not adequately capture the cyclicality of Cognex’s business selling industrial equipment.  However, it gets very close to the share price in the market.

FCF Valuation

Table 2 – Free Cash Flow Valuation of Cognex [Entries 2018-2031 Omitted for Clarity]

Discounted Economic Value Added Valuation

R&D should be capitalized in the firm.  This is the key asset which Cognex derives its revenues from.  Robotics factories tend to be singularly unimpressive and largely undifferentiated affairs.  The basis of the 21.3% return the Cognex has historically earned on its economic assets is largely the R&D.  As pointed out above, Cognex is probably not very efficient at managing its real GAAP assets.  My R&D capitalization schedule relies on assumptions, but I think reasonable ones based on my experience in the robotics industry.  These assumptions, along with the spread on employed economic capital, drive the value in the discounted economic value added method.  The spread that I used has to be pretty close to a fair estimate given the R&D depreciation method that I used, which assumes that R&D useful life is a random exponentially distributed variable with a mean of 12 years.

Discounted EVA Valuation

Table 3 – Discounted EVA Valuation of Cognex [Entries 2018-2031 Omitted for Clarity]

Long Form Economic Valuation

The long form model of the firm show in table 3 looks at the drivers of value.  As investment is variable over the period, I used the starting value of economic investment to .  This will likely understate the long form value of the firm slightly.  However, the long form appears to overstate the value of the firm compared to the other methods.  If an investment period of 10 years is used, the long form comes much more into harmony with current prices and the other methods.

Long Form Economic Value Drivers Model Table 4 – Long Form Valuation of Cognex

Conclusion

I’m not very enamored of public equity investing so I’m a little foggy on what the analyst terms mean.  In recent periods it has seemed like analyst terms like, “strong buy” and “buy,” mean things quite contrary to their common meaning—perhaps closer to “Be careful” and “Call your broker with a sell order ASAP.”  Going by conservative assumptions derived from historical data of the last economic cycle, I got prices that were very close to, and bracketed, the market price of the stock.  Cognex would be reasonable to hold in a portfolio if you expect earn the market cost of capital on your portfolio.  There is upside potential, but there are also risks the current price.  All in all, it looks set to return the cost of capital for the foreseeable future.

There is power in being able to say what amount of economic capital you are employing—regardless of where the accountants hid it.  It also allows you to look at any company like it is a bank.  The firm takes in capital from whatever sources, and using it for purposes that earn a spread over the cost of capital, then returning the capital and pocketing the spread for the owners.  This uniformity of treatment, really gets at the heart of what is creating value in the firm.

However, I’m not sure that any of the methods of valuation adequately speak to what the real risk of this company is—which is that it needs its research to match the needs of its customers.  The dogs might not eat the dog food, or they might unexpectedly ask for seconds.  These changes in customer demand are going drive immense fluctuations in all the assumptions that financial forecasts make.  It is a messy and localized business, but fundamentally, this is what really creates the value.  Just doing R&D is not going to necessarily create value, true of any asset, but the matching problems are much more severe in R&D and the rate of economic return incorporates a lot implicit assumptions about how management will make the assets perform.

Appendix

Data and Estimates

Data and Estimates

Table 5–Data and Estimates

Full Calculation

Table 6 — Printable Full Discount Calculations

The Knowledge Economy Cash Anomaly: Part 1

This is part 1 in a three part series about why technology companies hold so much cash on their balance sheets.  Here are Part 2 and Part 3.

The academics disclaim knowledge of a definitive answer as to why companies in the knowledge economy hoard a such a disproportionate amount of cash.   The problem is that the chart below has two branches where our classical understanding would only expect one.

Robotics (Blue) is firmly in the knowledge economy, using very few real assets, but a disproportionate amount of financial assets, to finance the company.

Robotics (Blue) is firmly in the knowledge economy, using very few real assets, but a disproportionate amount of financial assets on the company’s balance sheet–just like software companies.

The Expectation

The companies that form a cluster heading up close to the Y-axis are the traditional economy companies.  They are everything from utilities to content companies to retailers–some of them quite high-tech.  Basically, they have the real assets that they need to their business and a little bit of cash and securities to get them through the shocks of the next couple of months.

This is what financial economists expect companies to look like:  orderly, well managed institutions that collect cash from operations and distribute the operating profits out to shareholders and debt holders.  Since these companies have access to relatively liquid and efficient capital markets, they have no need to hold onto cash.  Good investment projects can simply be financed through issuing new securities or retaining more future earnings.

Tech Companies

Robotics companies and tech companies on the other hand horde massive amounts of cash–spreading out along the X-axis in the chart above.  Many of these companies, already profitable, could forego revenues for over a year.  And, oddly enough, the most profitable and most successful of these companies hold the most financial assets.  Nobody quite understands why companies do this.  The previous discussion, Is a Dollar Worth a Dollar on a Tech Company’s Balance Sheet?, reviewed some of the arguments for and against the value of cash on a company’s balance sheet.

Holding Cash Is Usually Bad

Most investors feel that excess cash in the company is a temptation to value destroying misadventures by management.  Particularly if management has incentives to grow gross profit, management can grow gross profit by deploying the company’s cash in less than profitable ways.  The classic example of this overpriced acquisitions.  Say you were the CEO of HP and you wanted to grow profits.  You might have heard about this company called Autonomy.   So you decide to buy it at market price plus a huge control premium.  Your profits go up, because you have HP + Autonomy’s profits together.  You get a bonus.  But your shareholders get hosed.

If the shareholders wanted to buy Autonomy, they could have owned it without paying the control premium.  Unless these so-called synergies show-up (and synergies are what go up the banker’s chimney after Santa Claus comes down–see I learned something in business school),  there is no reason to pay the control premium.   The control premium just gets pocketed by the previous owners and the bankers with all that value lost to the shareholders of the acquirer–those are your shareholders.

Conversely, if the company disgorges the cash, and you and your management team go to the market to raise new debt or equity to finance the purchase of Autonomy, Instagram, or any of another thousand bad acquisitions–the financial market has a chance to tell you that this is a really bad idea.  But if it is a good acquisition, the market will easily provide you with the money.  So all in all, investors tend to discount cash on the balance sheet and reward paying it out where they can reinvest it.  So why would companies hold all this cash?

Concentrated Ownership

Many tech companies are owned or controlled largely by single individuals or small groups.  The company represents a substantial portion–if not substantially all–of the wealth of the these founders.  Since they control the company, they are willing to take steps to decrease the risk to the company that are not economically maximal to diversified shareholders.

Consider this hypothetical:  Google and Apple are locked in winner take all product war for a small market that is worth $2Bn in market capitalization to Google now, but will go entirely to Apple in year unless Google spends $3Bn.  If all of your wealth is in Google and you couldn’t easilty get it out, you might be willing to have Google spend $3Bn to save $2Bn in wealth.  Your loss is now $1Bn instead of $2Bn.  However, if you are a diversified investor and own both Google and Apple, you want Google to let the business go and refund you the $3Bn.  You still have your share of the $2Bn in your portfolio and the chance to invest your share of the $3Bn somewhere else to earn a return.

Founder Payout Diversified Common Shareholder Payout
Spend money to protect Failing Business -1 -1
Do Not Spend Money to Protect Failing Business and Payout Cash -2 +3Bn

Less sinister, the company may just be conducting tax free diversification on behalf of the founder.  The effective corporate tax rate is below the individual tax rate, especially on capital gains.  While this is tax efficient for a founder, it may not be tax efficient for other investors.  The harm is probably not as stark as the example above, but it does raise the question about who the firm is being run for and rubs our Anglo-Saxon sensibilities about the primacy of the shareholders the wrong way.

Defense

The story here is that only Apple or Microsoft would ever even think about entering search knowing that Google has the largest market share, the best technology, and is sitting on $45Bn in cash.  If you want to take search from Google, you are playing a long game and an extremely expensive game.  It will fight hard and it has the resources to do this.  Potential competition is scared off, increasing the ability of the company to earn rents in on its primary operations.   While closely related to concentrated ownership story, this is actually favorable to the common shareholder if this is true.   Services like Siri, Wolfram Alpha, and IBM’s Watson cast some doubt on this story, but perhaps at least in the example of general consumer search it is mostly true.

The defensive effect need not be 100% effective to be worthwhile.  This effect is an extra return on the cash that shows up in operations, not financing, because of accounting rules.   Additionally, the company always has the cash, so there is option value.  In our example, if Bing every really started to rule search, Google could decide not to fight, and either sell or wind down search operations.  They still  have $45Bn to distribute even if the value of operations falls to zero, but the option to fight is inherently valuable.  With the cash to execute this option, it becomes more valuable, or credible.

to be continued…   Next up, option value of IP and distress costs

Jump to Part 2

Is anyone surprised the FAA is delaying UAS test site selection indefinitely?

I have to agree completely with the sentiments of Congressman Austria on this issue. The FAA is just dragging its feet.  The point of the test sites is to solve the issues of safety and privacy.  If these issues were completely worked out, we wouldn’t need test sites–do not pass go, do not collect more appropriations, proceed directly to airspace integration.

The point of the test sites is to work on these issues and give the general, civil, and commercial aviation community time to come to grips that some new craft are going to be joining their previously exclusive community.  Delaying the test site selection is the complete wrong approach.  The right approach is to begin testing–as most other developed countries already have.

How is privacy even the FAA’s jurisdiction?  In all seriousness, I hope that whatever regulations apply to UAS apply to cellphones.  I’m a lot more likely to have my privacy invaded through cell phone than through unmanned aircraft.

Even the Navy Has Market Risk (they just call it something else)

My long awaited article in Proceedings just came out!  You might not have been waiting for it, but I have!  I started the article over a year ago.   It was a slog.  I can’t quite believe that I’m now signing up to publish an academic paper on the capital structure of robotics companies.

Image Credit: DTIC

In summary, the U.S. Navy is making a terrible mistake in it unmanned maritime vehicle policy.  The Navy is basically designing all their programs for robots that swim in the water to fail.  The technology exists today to make really cool, useful maritime robots.  However, the technology does not exist today to build the Navy’s dream robots.  (Especially since the Navy’s secret dream is to need more manned ships of the type we have today.)  Essentially, the Navy is pulling the equivalent of refusing to look at Roomba because it is not Rosie.

I’ll try and expand upon some of the key ideas from the paper over the next couple of weeks.  Readers of this blog will be familiar with the core ideas which have been translated from business to military jargon.  The Navy has to figure out what it needs its robots to do and the ecosystem around them at the same time that it is working on building the systems.  That’s what we in business call market risk!  The Navy needs to take some steps to reduce that risk.  Although the defense acquisition process stacks the deck against the Navy and it has some truly heroic individuals working on the problem, as an institution, the Navy really isn’t putting forth an adequate showing considering we’re talking about the institution’s future.

As a patriotic citizen of the United States–and as someone who understands that the U.S. Navy as much any institution on the planet has guaranteed an era of global trade, peace, and freedom–I really want our Navy to have a bright future.  Everyone who studies the naval budget–horses and bayonets snark aside–knows that the Navy isn’t on a sustainable path.  Robotics offer the Navy a future even brighter than the past.  To have this future, the Navy will have to learn how to manage and implement this technology.  It won’t be easy, but there are solid principles for doing this.

P.S.

Also, readers, I want to thank you.  Thank you for being patient with a terrible layout, a casual tone, mixed quantitative/qualitative arguments, poor citation, and irregular tables. I do want you to know that you are reading a blog whose underlying ideas are good enough to go through peer review.  I, for one, commend you for that.  I hope that the ideas have a practical impact in advancing robotics that improve the world.  Now, stop indulging my self-congratulation and get back to putting more robots into the world!

Waiting for the relational database of robotics

If I were smarter and technically inclined, I would be working on a generalized solution to the problem of material handling.  Intuitively, it sounds really simple.  The basic task is to move objects in accordance with a larger process from one work or storage point to another.  Humans hate it, it is boring, sometimes dangerous, and doesn’t use much of the human’s ability most of the time.  Even more, it satisfies my test of a great robotics application (the Morris Test?):  By having the robot perform steps in a process can you favorably reorganize the larger process to change its key constraints?

Material Handling Robot at a Solar Cell Plant
Image Credit: Energy.gov

Why does it feel like we’re stuck in the 70’s compared to computing?!  What’s going on?  Material handling is to robotics what databases are to computers.  Certainly material handling is not the only application of robotics, but it is definitely one of the leading business applications of robotics, just like data bases in computing.  The need has all kinds of variation and tons of different degrees of depth and similar sorts of need across a huge variety of industries.  However, unlike the relational database, robotic material handling does not transfer to across industrial verticals.

The database industry was like this before the relational database.  Back in before the 80’s, every industry and often specific companies had their own database structure.  The organization was hierarchical and specific to the company or industry (i.e. arbitrary from a programmer’s perspective).   In order to build code or run queries it required elaborate local knowledge of how the data was arranged.  This prevented the transfer of technology between industries and made it difficult to build large software companies since work was largely non-transferable between clients.  The relational database changed all this.  It made it possible to transfer code and innovations between industries, allowed the deployment of programmers across industry lines, and built large dedicated software companies who were not system integrators.

All of the big industrial robotics manufacturers have a stake in material handling for manufacturing applications.  Additionally, there are many systems that we do not call robots by convention, such as optical sorting lines, which truly are robots in this space.  Beyond those solutions, several of the emerging manufacturers have stakes in material handling.  For example, Kiva Systems (Amazon), Seegrid, and Adept all have material handling solutions.  These manufacturers have solutions that serve what is essentially a single industrial niche.  If your environment or the need is different than the original application you’re out of luck.  Unfortunately, robotics manufacturers are not service businesses, a single vertical focus is a bad thing.

My initial take on this problem is that there are three problems that are closely related that all need to get solved jointly.  However, there may be value in solving the problems individually even before we realize the full benefit of the integrated solution.

First, is the problem of sensing, processing, and reacting to the environment to recognize the correct material and move it to the correct place.  My personal take is that this problem is close to solved as a stand alone problem, but that there is still some issues in integration with current solutions to the other problems.

Second, is the problem of interfacing with the larger process.  Robots that are unsupervised are notoriously difficult to synchronize with the larger processes.  I believe teleoperation in medical and field robotics is popular, not because autonomy cannot be programmed to complete the task at hand, but because the autonomy cannot be synchronized with the larger process.  This may have to do more with irrational flesh sacks running the larger process than any fault of the machine, but nonetheless, in integrated processes synchronization is key.  Note that the main differentiation between the robotic material handling systems discussed above is the method that they use to communicate with the larger process.

Third, we have not solved the mechanical generalization issue.  This will be the tough.  There are some things that require a forklift and some things that you cannot get a forklift too.  A forklift operator can get out and walk, but humanoid robots are not the answer.  Beyond revulsion with robotic anthropomorphism, there is no way that the human water bag is optimally configured for just about anything, let alone moving stuff around.  I was pretty unhappy carrying half my body weight in the Army, but for a machine that just pathetic.  Even if this problem could be generalized into a few basic types, that would go allowing the first two kinds of solutions to control the hardware.  The market as much as engineering may help solve this problem.  It will take deep pocketed speculators and market organizers to get this to happen though.

I don’t have a brilliant technical idea.  Further, this market probably will not have the same degree of network effects that power the database market (however there will be more winners).  That said, it seems like this problem has to be able to be generalized in a way that has previously eluded us.   When it does, hallelujah, we will finally have a piece of robotics technology infrastructure that will serve widespread adoption of a host of robotic applications.   The Jetsons will finally start to look backward!

Do you know anyone thinking about the future of aviation?

If you do, please make an introduction for me.

I’ve been thinking a lot about the future of aviation lately.  I’m trying to write a major piece for Patrick Egan at sUAS News and also thinking about this for reasons related to my business.  I’m not sure that we in the unmanned aviation community have done enough to think about what the future of the aviation industry is like.  Clayton Christensen’s Seeing What’s Next has a great discussion of disruption in aviation, but even though it was written in 2004, it makes nary a mention of unmanned aircraft.  Steve Morris at MLB Company also was kind enough to have lunch with me last week and talk about what he sees coming.

Photo Credit: DARPA / DTIC.mil

Hypothesized Developments in Aviation from Unmanned Aircraft:

-Aircraft building, particularly on the low end will approach a commodity industry more analogous to PCs or cellphones than current aircraft building paradigms.

-Unmanned aircraft companies (both builders and operators) are going to look more like software or networking companies than they are going to look like industrial companies, this has implications for both human resource practices and the capital structure of the companies.

-Scheduling, routing, and planning will be done according to the new paradigm.  Currently in aviation, everything is optimized around getting the most out of any particular flight hour or unit of plane time.  Unmanned flips this on its head and allows for the aircraft to be treated like other tools that wait on the main job.  Don’t know when you’ll need the plane up?  That’s okay, we’ll park it in the sky (maybe doing a lower value mission) until you need it.  Want to go from point A to B?  Great we’ll take you there, directly, when you want to go.  We will not worry about crew duty cycles, hubs, or returning the plane to its home base.

-Large airports will loose their centrality to the system–this is not to say they will experience a decline in traffic, but rather, they will not be the key limits on a network-like system of small airfields and ad hoc landing or operating sites (think more like a heliport than an airport).

Predicted Market Effects:

-Differentiation and customization will likely become the norm in unmanned aircraft operations.  Most airlines are pretty undifferentiated, but when the business customer is going to tie their ERP system to their aerial service provider’s dispatch system and automatically task aerial missions based on orders, sustained relationships and differentiated services are going to be much more meaningful.

-Data gathering / reconnaissance is likely to switch almost entirely to unmanned systems after the FAA changes the rules.

-Air Cargo is going to be significantly changed, mostly at the interface between trucking and air, with more work being done by air and less by trucking.

-In the long run personalized aviation, whether that is passenger aviation or other types of aviation consumption is going to be the big development.  Aircraft of today are like mainframes of the 70’s.  Only anointed experts who get to go into the restricted area can operate these machines.  Unmanned aircraft are going to be like PC’s, so cheap and easy to use that anyone can have one.  The possibilities here are quite remarkable.  Data collection, aerial work, cargo, and passenger transport are likely to feel the effects of this shift.

-Long haul, passenger, mass transportation will be the last segment to be effected.  The first segments to be effected will be small, light-weight, short duration applications.

So what else?  

I don’t really have a clear idea of how this effects incumbents.  It will definitely be change.  On the one hand, I think that the big guys at the top of the market will be fine.  I don’t expect Boeing or the airlines to disappear.  On the other hand, I don’t think that axis will have the control over aviation that they do today.  They will be more like bus companies and builders in the large automotive industry.

The cult of the pilot will be diminished (as it already is in military aviation) and air travel will continue to be democratized.  I believe that we are witnessing something akin to the introduction of the automobile.  Prior to the automobile, mechanized transportation had been too expensive and hard to use for anyone that was not an expert.  Prior to aerial automation, aircraft were too expensive and hard to use for anyone but an expert.  That’s changing, if we can hurry up the FAA, we have an amazing industrial explosion ahead of us.