hotbike

Not often do I post here in the VC forum, but sometimes I think about other means of traffic control. Right now, people are working on cars that drive themselves, and GPS systems that will let drivers know when another vehicle is in proximity. Actually, Ford is working on a system that will use GPS and WiFi , to let vehicles communicate with each other.
I want the reader to think of Air Traffic Control for a minute. Except for Helicopters, aircraft do not stop in mid-air. I've always thought of a bicycle being like an airplane, because it is lightweight, and the Wright Brothers were Bicycle Mechanics. Motorists may think cars are better than bikes, because , if you're not paying attention, a bike can fall and hit the ground. But that's not a valid argument , because Airplanes will also fall and hit the ground, if the Pilot is not paying attention.

But today, I want to mention a few things about Railway Signals. You must be familiar with Traffic Lights, Red=STOP, Yellow=Caution, Green= GO. We know the DMV tests Drivers to make sure they know these signals , but what if lights had other aspects?
I will list a few from this site:
https://www.lundsten.dk/us_signaling/...995/index.html

Instead of just stop, caution and go, (three), Railway Engineers, Train Drivers need to know a few more additional signals.
Here they are, from the Santa Fe Railway:

Clear
Approach Limited
Advance Approach
Approach Medium
Approach Restricting
Approach
Diverging Clear
Diverging Approach
Restricting
Stop and Proceed
Stop
Distant Signal Approach

Additionally, Union Pacific has a few more:

Diverging Clear Limited
Diverging Clear Slow
Diverging Approach Limited
Diverging Approach Slow

Forget about driving . Most drivers are challenged enough by the three different signals. If they had to learn Seventeen different signals, the DMV wouldn't be issuing as many drivers Licenses as they do today.

I would feel safer with Robots driving the cars, provided they use GPS and WiFi, and follow at least as many signal aspects as the railroads do.

Note: the railroads traditionally send electric currents through the rails, which are activated when the presence of a Train short-circuits the rails. GPS is capable of keeping track of cars, the way railroads keep track of trains. There are also CTC, or Centralized Traffic Control, and ATS, or Automatic Train Stop on some railroads.

Anyway, I think it's coming to a Future near you.

I kind of like the idea of testing drivers for all Seventeen signals, not just the Three that they can't even remember now. It would keep them off the roads.

I'll try to elaborate on these ideas more, in the coming weeks, as I follow this thread.

But now I welcome your input, please feel free to reply. Don't embarrass yourself by saying it will never happen, or I'll come up with a webpage or two showing it's already been done.

dynodonn 

I'm not sure if I would feel safer with that thought aspect, especially with the system being prone to cyber attacks/hacking, even the best security systems can still be breached given enough time, energy, and willingness of an individual or a highly trained and coordinated group.

wphamilton

I tend to agree that it's inevitable but I think that there has to be a distributed collision avoidance system, rather than centralized. For example, each vehicle having the ability to transmit location and velocity, receive and process those transmissions and react accordingly. I also envision detection of passive objects (which will not transmit) - radar, sound, or some other technology - as even more critical to the system. We'd never accept robotic cars driving blindly into unexpected obstructions.

It brings up mental images of cycling in traffic, and all the vehicles automatically slowing and traffic swirling around us. Now THAT would be turning the tables, traffic being automatically forced to adjust to us rather than them whimsically forcing us to the gutter, or emergency stops or other evasion.

dynodonn 

My concern is what "fail safe" safeguards will there be if one vehicle or a fleet of vehicles malfunction, will the vehicles pullover and quietly park on the side of the roadway, slow to a stop in the middle of the roadway blocking traffic being a safety hazard, or just careen out of control like a motorist losing consciousness.

hotbike

Approach Limited- Proceed prepared to pass next signal not exceeding 60 MPH and to advance on diverging route.

Advance Approach- Proceed prepared to pass next signal not exceeding 50 MPH and to advance on diverging route.

Approach Medium- Proceed; approach next signal not exceeding 40 MPH and be prepared to enter diverging route at prescribed speed.

Approach Restricting- Proceed prepared to pass next signal at restricted speed.

Approach- Proceed prepared to stop at next signal, trains exceeding 40 MPH immediately reduce to that speed.

Diverging Clear- Proceed on diverging route not exceeding prescribed speed through turnout.

Diverging Approach- Proceed through diverging route; prescribed speed through turnout; approach next signal preparing to stop, if exceeding 40 MPH immediately reduce to that speed.

Restricting - Proceed at restricted speed.

Stop and Proceed- Stop, then proceed at restricted speed.

+++++++++++++++++++++++++++++++++++++++++

I'm not just suggesting that cars be able to drive themselves, that's inevitable. But for drivers who still want to control their own cars, we should have better signals, based on what the railroads use. The railroads have more signals, in between red and green, instead of just a simple yellow. I've amended the list with what the signals require the driver to do, (ABOVE).

The Engineers who drive the trains are only human, but they know the signals. I don't think it's unreasonable to expect motorists to adopt a new signal system. The railroad signals also say what speed...to travel, all the way to the next signal.

Here's a link to the NORAC rules:
https://broadway.pennsyrr.com/Rail/Si..._us_norac.html

I foresee something like railroad CAB-Signals coming into automotive use, perhaps a GPS or WiFi Based device, which will sit on the dashboard of the Automobile, and tell the driver not only whether to stop or go, but at what speed, all the way to the next signal, or Block.

Once a system like this is in place, it would be much easier for robots to drive cars.

Speeds might have to be adjusted downwards from the railroads system (ABOVE), since trains do travel faster than cars.

I'll sum it up thus:

"Red, Green, and Fifteen Shades of Yellow".

CB HI

How and at what age do you plan on implanting the gps - wifi units into pedestrians to keep them from being run over.

dynodonn 

Just more distractions to the already distracted motorist. I'm happy when motorists just follow the existing traffic laws that are already in place.

Stealthammer

This was actually one of the "selling point" for Blue Tooth early on. Since Blue Tooth is rather limited in its range, the idea was that cars would "scan" for other obstacles and either warn the driver or direct the car to avoid the obstacle. It was kind of a scary thought when I first heard about it, but noting the limited driving skill of many drivers, it could save a lot of cyclists lives. Another interesting concept I heard about was that when your car pulled into a parking lot it would scan for empty spaces then alert you where they were. This really sounded like a great idea and I hope that someone actually gets it to work....

sggoodri

I believe that by the time autonomous vehicles are practical enough to be productized, they will all be able to detect and track pedestrians and bicyclists in the streetscape using non-transponder sensors such as visible/IR video analytics and laser radar. Skeptics should consider the XBox Kinect, a relatively inexpensive consumer device that can not only measure the presence of a human but their 3D body pose and motion. Extending the range of this sort of 3D sensing capability is relatively straightforward and there are a variety of maturing technologies that can facilitate it at costs that continue to decline. I did my graduate research on computer vision and autonomous vehicle control, spending a lot of time evaluating systems for avoidance of dynamically moving objects, and my current work involves a variety of video sensing technologies including night vision. Seeing and tracking people really isn't difficult for a machine to do, even in the dark. What's more difficult is predicting changes in a person's movement, e.g. is a pedestrian going to step off the curb in front of the vehicle. A bicyclist following the normal rules of the road is far more predictable than other people that an autonomous driving system will need to accommodate.

dynodonn 

Again, my question is what happens when there's a malfunction, what safe guards will be in place. Also, like a lot of vehicles they're fine when new, but as they age they tend to fall into disrepair especially when the owner cannot keep in step with the upkeep.

prathmann

Robot-driven cars are already in use today. Google has modified a number of cars to be completely computer controlled with the driver acting only as an observer and able to take control in an emergency to satisfy legal requirements. AFAIK, only Nevada has so far granted authorization for use of these cars on public roads, but they are also being used in more controlled campus settings and in authorized trials.

There are also other similar research efforts underway elsewhere. Here's a clip of an autonomous car from the Tech. Univ. of Braunschweig driving in city traffic (this is in Germany):
https://www.youtube.com/watch?v=BrmorE5W1tM

Personally, I would feel far more confident that these computer operated cars would pass me safely, avoid right and left hooks, etc. than I am with human operated vehicles in my proximity as I cycle.

wphamilton

That's the same thing I brought up "I also envision detection of passive objects (which will not transmit) - radar, sound, or some other technology - as even more critical to the system. We'd never accept robotic cars driving blindly into unexpected obstructions." Automated driving is not feasible in general unless that capability is 100%. The alternative or intermediate scenario is having automation on high-speed throughways which is not engaged on surface streets, which implies restricted access to "robot car" streets.

genec

Well of course the vehicle will take itself in for service, and if it is not maintained, it will simply find the nearest cliff and drive off...

Most likely if the system does not pass power on self tests, it will not go.

sggoodri

The trouble is, robot-only streets can't happen until the number of robot car owners hits critical mass to pay for them, and people won't buy robot cars without somewhere to drive them. That means robot and human drivers will probably be mixing on existing roadways. And because some errors/traffic violations by human drivers can create crash scenarios that robot drivers cannot avoid due to the limitations of vehicle maneuverability, these systems will not be 100%. But lawfully operating bicyclists will not create any significant difficulty for robot drivers, and so there is no safety justification for removing lawfully operating bicyclists from streets that include a mixture of human and robot drivers.

sggoodri

When I took the NCSU robotics lab's 400+ pound autonomous robot to demonstration events to show it driving through unmapped, dynamic environments without running over people, the first thing people would do is jump in front of it to watch it slam on its brakes. Fortunately, I had designed the collision avoidance system with redundant proximity detection systems and adequate decelleration behaviors to accommodate this, but I found the public's confidence a bit unnerving. I fully expect some bicyclists to recognize robot cars and not only ride more assertively, but even bully them around in ways they wouldn't dare act with human drivers.

genec

I can't help but wonder if this issue of lawful robot cars vrs "less than lawful" human drivers will cause any sort of shift in the blame of humans in collision situations? Everybody wants to assume the robots will fail, but if over time it can be shown that the robot drivers are less likely to be the cause of collisions, there might be some rather interesting repercussions toward human drivers.

sggoodri

Possibly, but the public's love/hate relationship with automation technology manufacturers will probably result in the legal requirement that robot drivers record all of their sensor and actuation data and provide it to crash investigators. That data should make determination of fault a more objective, scientific process.

prathmann

Certainly, and I expect that the result will be to provide conclusive evidence that fault almost always lies with the human drivers. Once human-driven cars become a distinct minority there may well be a shift away from judicial leniency for crashes that are the result of momentary lapses of attention or judgement. There will be the feeling that the at-fault driver should really have chosen to let a robot do the driving. And judges and jury members (at least those with a newer car) will no longer have the empathetic feeling for the defendant that they could someday make a similar mistake.

wphamilton

Very true. The only way I see to evolve into that (robot-only roads) would be something like the express toll roads, accepting a transponder signal rather than quarters. But that would have so many limitations I don't see it gaining traction. So the practical option to pursue is the first one, 100% avoidance capability.

Clearly a person or vehicle jinking into the lane immediately in front of a robot could never be 100% avoided. Yet, I don't see that as a necessary requirement. I think that the detection technology is sufficient already, and programming the judgment is the challenge. Granted that is a big challenge but it could be far safer for the other guy. Maybe. I don't know honestly. There's always exceptional situations that may require human judgment. But with the advances in expert systems AI I'm leaning toward the computer being able to handle it.

Here's a for-instance just from a recent development. A group successfully designed an algorithm and working system to anticipate red-light runners by observing them just prior to the intersection. They were something like 85% accurate which to me is impressive. That's good enough for their application, flashing a warning on someone's dashboard, but not good enough for a robotic driver. Like you imply, it needs to be ninety nine point something. Not only for running reds, but the even more difficult problem of lane changes, sudden stops and accelerations, merging and so on. Not to mention planning the "out" as we're taught. It's a hugely challenging problem so you may be right.

sggoodri

In the wealthy United States, I can imagine robot drivers becoming the majority of vehicles driven. I have a hard time imagining it in poorer developing nations; I think mixed traffic will be around for a very long time. I suspect that there may be economic and regulatory pressure to eliminate low end manually piloted motor vehicles in favor of nonmotorized vehicles for those who cannot afford the more expensive robot vehicles. In the developing world, there will be so many people unable to afford robot cars that I think there will be a strong political force demanding that robot cars operate safely around non-motorized users, at least overseas.

genec

I have long suspected that a shift to "mandated" robot cars would be done on a fuel demand basis... where the robot cars use some unique fuels that the currently human driven cars do not use. Over time, gasoline and gas stations would be eliminated, with the result being that in order to have an operating vehicle, one has to embrace the new fuel. Gas will still be available, but not ubiquitously as it now is; thus forcing human drivers to endure quite a bit of inconvenience to continue driving their "dirty" vehicles.

No doubt used American gasoline powered vehicles will be come quite a commodity in developing nations.

Doohickie

tr;dl

(too rambling. didn't read)

genec

The main problem with the OP thinking and idea is the fact that strictly VC cyclists would also have to memorize the "15 levels of traffic signals" too, thus his idea eliminates all human controlled transportation completely.

On the other hand, robot driven vehicles WILL adhere strictly to the rules of the road and thus will actually encourage strict vehicular cycling. The VCers will finally get what they have often touted, a system based fully on the logic of the rules of the road. (of course all this time the missing factor was the fact that human motorists tend to "invent" new rules as they go... )

dynodonn 

That is fine on the vehicle's initial start up, but my main concern still about if a malfunction occurs during normal operations.

"Redundancy" is definitely the major key word I would use when it came to the use of autonomous cars, along with aircraft type certification/maintenance.

genec

While I do fully agree... I have to chuckle at the demand that computer controlled cars have redundancy... we certainly don't demand any of the above from human driven cars.