The previous six National Bus Trader articles on this subject stabbed at some highlights and low-lights within the extraordinary spectrum of socio-economic, institutional and other issues encompassed by our transition from humanoid-driven to robotic vehicles. At this point, I thought it might be helpful to take a quick glance at some of the hardware that serves as the robots’ mechanical fixtures, apart from the electronics and the digitalia: Cameras and sensors. These components were employed in “transitional” or “steppingstone” efforts along the path to truly driverless vehicles. So I feel it is worth a look at how these technologies were used and abused at this earlier stage of HAV (highly-automated vehicle) development. Should the reader wish to view the math in the robots’ brains, I recommend Multiple View Geometry in Computer Vision by Richard Hartley and Andrew Zisserman. The bible for artificial intelligence. Way over my head. If also over yours, no apologies necessary.
One thing that has always saddened me about often brilliant technologies has been the bungling, over-extravagant and often dysfunctional misunderstanding that accompanies their application and use. Benjamin Franklin and Thomas Edison must be spinning in their graves from all the usages to which harnessed electricity has been put: Gaudy ornamental lighting (e.g., neon signs), obscene inventions (e.g., sex toys), slothful products (e.g., electric toothbrushes) — along with the plodding, inefficient infrastructure (e.g., dams) constructed to generate it. Unfortunately, we live in the Age of Instant Tea. So Messrs. Franklin and Edison will likely continue spinning.
One victim of these trends is among my favorite suppliers, Rosco Mirrors. Along with other excellent mirror manufacturers (e.g., Mirrorlite and Tiger Mirrors) which have also expanded to also become camera and sensor manufacturers, Rosco and its competitors have created one genuine safety improvement after another. We are all familiar with some of this sub-sector’s innovative safety products — like the parabolic “banana” mirrors that quickly became regulatory fixture on, and a requirement for, schoolbuses in the 1980s. Yet our industry’s adoption of the best of this technology has zigzagged, if it has not hiccupped. School bus mirror systems are significantly superior to those on other passenger transportation vehicles. Even a schoolbus’ plain-old convex mirrors are larger. So why have we not installed them on other buses and trucks?
More recently, Rosco in particular, has invested considerable resources as one of our nation’s pioneers in the development of semi-autonomous technologies for buses, motorcoaches and trucks. To the degree these vehicles help keep us safe, we owe Rosco and its competitors a genuine debt of gratitude. And we should have a deep respect for their product-development excellence. At the same time, these suppliers and their employees must be deeply saddened by the ineptitude and misdirection displayed by certain public transportation agencies in the misapplication of their products to serve really stupid purposes to reach goals we have met decades ago.
This revelation is particularly sad in the context of the uber-intelligent use of these same technologies (and much more) in aircraft and ships. They allow pilots to take breaks during routine segments of flight — leaving them more refreshed for landings and take-offs. But far beyond improved safety, these technologies are showing us things in outer space and at the bottom of the sea that, until recently, we could never have dreamt existed. Every common fisherman understands the challenge of keeping a boat or ship still in a moving current, much less in an ocean of moving water. Yet the robots are coming amazingly close to accomplishing this with huge marine craft. And modern aircraft, particularly its military applications (including drones), would be impossible with a considerable volume of artificial intelligence. Even still, before launching a strike based on drone or other aircraft or satellite surveillance, a human being always views the footage, considers multiple factors (on- and off-camera) and makes the decision. This is not the case with driverless vehicles.
Crafting with Crayons: The Schoolbus and Transit Communities
Using only the really thick red, blue, green and yellow crayons we were given in kindergarten (because we would likely break the thinner ones), some of our schoolbus and transit communities have literally abused the potential with which many suppliers’ cameras and sensors endowed them. Many at the rear of this development actually use the terms ‘camera’ and ‘sensor’ interchangeably. Otherwise, some of the things done even at the semi-autonomous stage of commercial vehicle development in this field are disturbing.
- I cannot understand the gall of characterizing the Student Detection System as a “new” safety feature (WABC, Plainsboro, NJ, January 14, 2016). Or the arrogance to practically plagiarize from a child’s first admonition about crossing a street – “Stop, Look and Listen” — by claiming that, “This will be another sound we will get used to. Stop, pause and look.” I clearly remember seeing VORAD’s motion-sensing technology prominently displayed at trade shows, and written about and advertised in trade magazines, at least 15 years ago. I just do not recall many OEMs or convertors offering it as standard or optional equipment.
- I cannot fathom the marginal accomplishment of detecting any movement within 10 feet of the side of the bus — as one sensor manufacturer noted its technology does — boasting about the curb-side band of vision that, “… is … critical because this is where the children are getting on and off.” Perhaps someone can explain to this individual that, the width of the driver’s view along the curb-side of a full-size school bus at the curb-side rear tire (through either the flat or convex curb-side mirror) is wider than this distance.
- Further exaggeration about sensor technology is nonsense: “Throw mud on it and its still working.” What is still working? Unless the lens can wipe itself off, melt or vaporize the mud, this is subterfuge, if not an outright delusion. And even if this device could still sense something, how could it make any meaningful distinctions about it?
- On January 11, 2017, the New York City Transit Authority issued a press release about its evaluation of a Pedestrian Turn Warning, Collision Avoidance System, designed by another excellent supplier with whose products we are well-familiar: Clever Devices. According to the NYCTA, Clever Devices’ new surprise is a “smart sensor-based technology designed to prevent forward and side collisions by alerting the bus operator with visual and auditory warnings.” Is the NYCTA referring to its visually-impaired bus-operating work force? If drivers cannot hear and see with mirrors and cameras designed decades ago, much less a clean windshield kept clean by wipers and washers, what is the “smart sensor based technology” possibly going to do for them? If drivers are failing at these tasks, the solution is to monitor, terminate and replace them.
- In a December 2015 another bus magazine covered this same transit agency’s 60-day pilot program testing a “Shield System.” This system ostensibly “…alerted drivers based on the location of a pedestrian and [the] severity of the threat” — a technology that this transit agency claimed could “filter out” pedestrian proximity that is non-threatening.” IN fairness, one advantage that this technology has is that it can view multiple positions, and multiple phenomena, simultaneously — whereas a driver can only sense or observe one at a time. Yet instead of noting this, the NYCTA actually claims this technology could “lock in and follow pedestrians and their course if they are deemed to be collision-likely.” Here is a new term for you: “Collision-likely.” To do this while mounted somewhere on a bus, the bus would have to drive up on the sidewalk, cruise down narrow alleys, climb steps and occasionally board a subway. Are you kidding? Do drivers not have windshields? And eyeballs? Or are they too pasted with mud?
Frankly, I hardly think New York City is a good venue for testing any moving vehicles with any technology. A 2009 study by the NYC Department of Transportation examined 7,000 vehicle-pedestrian accidents in Manhattan alone during a period beginning a few years earlier. The study actually informed readers that their odds of crossing safely were increased by jaywalking across streets at midblock positions — yet never bothered to point out that this tactic might have some merit on slow-moving side-streets with a single lane of traffic moving in only one direction — as opposed to eight-lane, high-speed, bi-directional boulevards like West Side Highway! If nothing else, that study also demonstrates that vehicles are “collision-likely.”
NYC pedestrians act as though moving vehicles are a mirage. Many cross streets against the light with headphones on. So what is some loudspeaker on a bus going to tell them that they cannot see? Or is this technology’s auditory functions designed solely for pedestrians living in their phones? Are pedestrians living in their phones “collision-likely?” If an autonomous vehicle had to avoid them in a places like Manhattan, they would have trouble moving a millimeter. The traffic-reducing benefits of such vehicles from decreased following-distance seems like it would be more than offset by a degree of caution that would slow travel speeds to a half — unless the artificial intelligence develops to a point where it can read pedestrians’ and fellow-motorists’ minds. And even this might only help if all or most of these individuals even thought about where they were going.
Some of the braggadocio about camera and sensor technology recalls a poignant old political joke. Decades ago, when stiff, Kennedy-clansman Sargent Shriver was running for President in the Democratic primaries, he walked into a bar in a working-class Boston neighborhood and asked the bartender for a Courvoisier. The Barkeep replied, “You want mustard on it?”
Challenges and Chimps
Let us consider some rudimentary challenges — driver versus sensor/camera — challenges I do not suspect the robots can yet meet:
- Can a camera or sensor tell the difference between a pedestrian carrying a cell phone versus a hand grenade? More interestingly, with a bit of help from the internet, cannot a smart high school student transform his cell-phone into a bomb? As at least half of all terrorist attacks in the world focus on buses, one thinks that this capability should be important. Sensors detecting “some object” within 10 feet of the bus are not even steppingstones to such capabilities.
- Anyone who knows the difference between a “fixed” hub (a traffic light) and a switch (a traffic cop who can make choices) understands that the need for the camera/sensor to linger longer at certain images, or return to them more quickly and/or more often. But without an extraordinary database that sounds undoable and unaffordable, I doubt any camera- or sensor-based system can do this. However it is set to pivot or rotate, a camera or sensor will normally rotate or pivot with some degree of regularity. Or perhaps their pivots or rotations may be influenced by the proximity of certain objects and their movement. Yet movement changes constantly and unpredictably. Some things are worth more, quicker and longer looks than others. A camera or sensor can surely “keep its eyes moving” and “get the big picture.” But can it “expect the unexpected?”
Even if these devices eventually perform magic, think about the carnage during the transition through these intermediate technologies when drivers actually begin relying on these technologies.
I stopped writing this piece after citing malarkey from a handful of articles only because there was so much of it and National Bus Trader articles can be only so long. But if the reader would like to view good technology used badly, he or she should follow some of the news about interim or semiautonomous vehicle technologies.
As a veteran of at least 600 public transportation-related lawsuits, and 17 years writing about them for National Bus Trader, no one need tell me that many bus drivers do retain or apply their defensive driving training. But the better ones do. If we cared more and paid them better, and paid for more and better management, most of these drivers would at least have a chance. So while a sensor or camera does not have reaction time or experience fatigue, it also cannot make judgments. It is a hub. Distribute the wealth a bit better, put traffic cops back on the beat, add a dash of enforcement, and gridlock would fade away like fake news. So too would many transportation problems if we gave live Earthlings the chances they deserve.
While accepting the inevitable, many public transportation bureaucrats are abusing the otherwise-excellent developments of vehicle designers and engineers. They are doing a marginal job of thinking clearly about some of the early steps in the robots’ development. Some of the early bungling noted suggests that driverless vehicle might have spent more time on the test track before they were released into the general traffic stream.
I suppose it is false hope. But I surely hope that developers of driverless vehicle technology do a better job than their counterparts have done with its transitional predecessors. If not, there may soon be no need for terrorists. Our bungling misuse of technology will do the dirty work for them. As cartoonist Walt Kelly once proclaimed, “We have seen the enemy and it is us.”