Kevin Cameron
Everything will be just fine – or will it?
Autonomous vehicles are coming. But can they coexist with motorcycles?
My apprehension over what will happen as motorcyclists encounter autonomous vehicle traffic has been somewhat diverted by troubling events in aviation. Boeing’s 737 Max 8 single-aisle airliner has been grounded since March 18, 2019, while government agencies and Boeing engineers have worked to understand (1) how a modest bit of digital flight control software called MCAS apparently caused two crashes that killed everyone on board, and (2) how to now make the aircraft safe enough to return to service. A grounding lasting 15 months suggests that more may be involved than just a few lines of code. It has been claimed that Boeing has lost 18 billion US dollars in the first 12 months of the grounding. Big stakes.
Do you suppose that the software, sensors, and control apparatus required to implement self-driving cars and heavy vehicles will be more complicated, or less complicated, than a system designed to prevent a single airplane from pitching up enough to stall in particular circumstances? The pilots in the two crashes seem to have tried to assert human control over automated systems, but ran out of time.
If I am thinking about this, surely many engineers and planners working on the problems of autonomous cars and trucks are thinking about it, too.
On the one hand, we are assured that AI (Artificial Intelligence) is advancing so rapidly at present that by the time robot cars enter traffic in large numbers, all potential problems will have been 100% resolved. AI has been hailed as making possible ‘deep learning’ – the discovery of hitherto unrecognised patterns in big data through sheer computational speed, acquiring lifetimes of experience in seconds. This will, it is hoped, make it possible for the cloud-connected systems on vehicles to manage dense traffic, in cities or on motorways, in ways that no mortal human can live long enough even to imagine. Essentially this is an invitation to ‘Have faith in our all-knowing algorithms – everything will be just fine’.
Mashing data is one thing, but having a workable idea is another. In MotoGP, Yamaha have suffered premature rear tyre fade for four or five seasons now. Surely their engineers have diligently worked this problem, but so far it hasn’t budged. When MotoGP began in 2002, Yamahas were uncompetitive until distinguished engineer Masao Furusawa had an idea – that the unique inertia torques of their flat-crank in-line four engine were either interfering with tyre grip or making it harder for the rider to sense tyre behaviour. When his changes were implemented, MotoGP dominance shifted from Honda to Yamaha.
Without ideas, thousands of hours of engineering and computer simulation could just as well have been spent working crossword puzzles. Work without understanding brings no results.
I think I can lately hear some caution creeping into the assertions of the robot car lobby. After all, even the question of legal liability has yet to be answered.
If a future Level 5 robot car (no steering wheel, brake pedal, etc.) happens to identify a motorcyclist, stopped at a traffic light, as a flock of sparrows passes or there is a July snow flurry, who will be found responsible if the robot bangs into the bike? Without a driver it’s hard to claim human error or negligence. Pigeon product on a sensor? An MCAS-like software glitch? Something up in the cloud? Big execs may be popping awake at 4am, having dreamed that it’s their auto, or sensor, or software company rather than Boeing who has just lost 18 billion US on a bet that automated systems are now ready to take over public safety.
Of course, at least in the US, there is another way. When nuclear-powered electricity generating stations were first proposed, the insurance industry wanted nothing to do with them, so the Federal Government forthrightly announced via the Price-Anderson Act of 1957 that it would underwrite any misadventure that exceeded a certain maximum.
Given the natural behaviour of large organisations and the example of 15 months consumed by the re-consideration of the MCAS software, it might just be that we won’t have to sweat being crushed by robots for a fairly long time. Long before an actual steering committee can be convened to consider how best to define the problems, study groups will be required to gather, collate and interpret data. Then a series of meetings to set agendas for future meetings will be required. Once the problem has been better defined, it must be divided into functional areas, each of which is entrusted to its own committee. To prevent separate development of different systems resulting in major incompatibilities, periodic systems integration meetings must be held.
With all that happening, development of Level 5 ought to take decades. Let’s go ride.