Advanced Driver Assistance: Still Not Seeing Motorcycles

[advrider]

Here we all are, hurtling into the future. Electric cars, electric motorcycles, alternative fuels, smart devices, the Internet of Things. It’s all around us and creeping into our daily lives.

Soon the future will include self-driving cars. But how distant is that future? That’s the big hairy question, right there. The more computers and smart devices make their way into the vehicles we drive and share the roads with, the more the drivers in them stop paying attention to the task at hand. And the AI systems in cars are not good enough for the confidence too many drivers put in them.

The Connected Motorcycle Consortium

That, at least, is a possible conclusion reached by the authors and researchers from the Connected Motorcycle Consortium. They just released a white paper on “Powered Two Wheeler Conspicuity.” We all know becoming more conspicuous in traffic can help us avoid collisions. But the question is becoming, “more conspicuous to whom?” Or, more specifically, to what?

Teslas Aren’t The Whole Problem

You’ve probably been hearing about “self-driving” Teslas that have recently plowed into motorcycles, killing the riders. That’s definitely a problem, and FortNine released a video (below) recently telling us all about that particular problem. He makes a lot of excellent points as well as a few scary ones. The motorcyclists recently killed by Teslas (more specifically, Tesla drivers), apparently had low, dual taillights that the Tesla “autopilot” misconstrued as far-away car tail lights instead of close-up motorcycle tail lights.

The problems here are many. First, Tesla’s “auto pilot” is a misnomer. There’s some assistance, but Tesla is very clear that drivers should not fully rely upon the system. Drivers should always be alert and aware in the driver’s seat. The owner’s manual even states: “Navigate on Autopilot does not make driving autonomous. You must pay attention to the road, keep your hands on the steering yoke at all times, and remain aware of your navigation route.” Fat chance, eh?

Also, the fact that Tesla is removing their front-facing radar in favor of relying on cameras alone is also worrying. Cameras can be obstructed by rain, snow, mud, dead bugs, etc.

Adaptive Cruise Control Is Already Here

Let’s take Teslas out of the equation. Manufacturers have fitted adaptive cruise control and lane assist on a lot of modern cars. Though it has seen active use for some time, adaptive cruise control still, has a lot of issues “seeing” motorcycles. This is a much bigger problem, because so many more cars are outfitted with these systems than a more “Tesla-like” “auto pilot.”

The 46-page white paper from the Connected Motorcycle Consortium lays out their methods, limitations, recommendations for further research, and ideas for improvements. I highly recommend reading the whole thing, if you have the time and inclination. It is a moderately unnerving read.

A “Global Motorcyclist Target”

The researchers set up various scenarios with a stand-in for a motorcycle, called a “Global Motorcyclist Target.” This is basically something shaped like a motorcycle with a rider (blue pants, black shirt, helmet) to be used in their testing, instead of endangering an actual motorcyclist. The model represents a motorcycle without a windshield or front fairing, with a small (600cc) engine, in the middle of the lane. There’s a troubling factor: the systems relied on motorcycles traveling in the center of their lane; otherwise, they can be confused with adjacent traffic. Who among us habitually rides in the very center of a lane?

Advanced Driver Assistance Needs More Advancing

They used cars outfitted with what they refer to as Advanced Driver Assistance Systems (ADAS). Mostly, this is adaptive cruise control. Even with lots of sensors all over the equipped cars, the systems failed to recognize that GMT a distressing number of times.

The scenarios replicate real-world encounters we’ve all had with cars: overtaking them on a multi-lane road, stopping in front of them, slowing in front of them, turning left in front of them, being between them and something like a box truck. The cars’ systems were not reliable enough not to hit motorcycles in too many scenarios.

Fully 32 per cent of the time, a car’s ADAS failed to recognize a stopped lead motorcycle (or Powered Two Wheeler, PTW) in front of them to brake in time, in this study.

Humans, Computers, or Both? Or Neither?

The study detailed the way RADAR, LIDAR, and camera systems work in the ADAS systems, and what their limitations are. The human eye and brain might not be perfect, but they are much, much better at recognizing motorcycles in all kinds of weather than these systems are. Given that the driver is paying any attention at all.

This is not a new problem. Studies have been showing for years that systems like adaptive cruise control fail to see motorcycles too often. But as more and more cars come standard with “safety” systems like these, the motorcycling community will encounter users who rely on them too much, to our detriment. 

Read Your Owners Manual

All vehicles equipped with these systems have repeated warnings in their user manuals that the driver is still responsible for not hitting other road users and must stay alert at all times: hands on the wheel, mind in the moment. How many people do you think really read their car’s owner’s manual?

Driving Manufacturers to Improve?

We have always known that motorcyclists are among the most vulnerable road users. Studies like these can help to motivate manufacturers to make their systems see us more reliably. The trouble is, the systems are mostly tested on bicycles and pedestrians, which we all know are not generally found in road lanes and travelling at the same speed as the rest of traffic. The assumption that the system can see a pedestrian or a bicyclist, that it therefore must be able to see motorcyclists, is misguided at best. And yet, that’s where we are. 

If studies like this one are performed more often and with specific systems, perhaps they can be rated and therefore improved. Getting drivers to pay attention to what they’re doing is apparently, though, a non-starter.

A Different Kind Of Visibility

Instead of relying on ever-more disconnected drivers on the road to see us and keep us safe, it might be time to modify our motorcycles. Up to this point, we’ve tried bright colors, and interesting lighting, and sometimes even loud pipes, to make ourselves conspicuous to surrounding traffic. It might be time, instead, to think about playing to the AI as well as the human drivers.

This means that instead of trying to catch the human eye, we need to think about radar deflection, and camera lens reflectivity. Some of us ride motorcycles that have giant, flat panels (think hard ADV luggage) on our bikes. Those are great for visibility by radar and cameras. Those of us who ride naked bikes or motorcycles with soft luggage may instead be nearly invisible to those same systems. Motorcycles in general lack a large, flat surface that’s easy for radar to “see.”

Your luggage may make the back of your bike very visible, but what about the front? The dreaded left-turner might not “see” the bike coming toward them unless it has a barn-door windshield fitted to it. Otherwise, it’s a bunch of uneven, pointy surfaces; that is the most challenging radar target.

In An Unfortunate Technology Valley

Finally, we may simply currently be in a technology lull. Too many drivers are relying wholly on poorly-performing systems. Those systems are improving, but cars are still killing motorcyclists too often. One solution might eventually have to be vehicle-to-vehicle (V2V) communication systems. Once the majority of vehicles on the road can “see” each other by sensing, or reading, or hearing each other (whatever you’d like to call it) with a communication system, distracted drivers will stop killing other road users. The issue here, though, is privacy. Can you be tracked by your V2V communicator? Probably, just like you already are by your electronic toll reader.

Vezi sursa