A problem inherent in any automatic transportation system is how to avoid collisions, specifically a following vehicle hitting a leading vehicle from behind. Two factors that make PRT susceptible to rear-end collisions are the relatively tight spacing between vehicles and any autonomy given to these vehicles.
It’s my nature to want to push the envelope in terms of spacing, autonomy and speed, so naturally I’m trying to figure out how to do it safely.
Historically PRT designs have had a set line-speed. The only variation was to drop back or increment forward by a set amount, and to have a slower speed around off-line stations. This makes keeping vehicle spacing very easy, and simpler in terms of getting a product to market. With self-banking, hanging vehicles, however, the only limit to the speed is what the passenger can handle. If a uniform speed is used, it must be one that is comfortable for passengers who are older, or less adventurous, or prone to motion sickness. The whole system gets slowed down for a small percent of potential passengers. If we are serious about creating a system that can make a profit and also compete with cars, I think we need to go as fast as is practical and comfortable for each passenger who is traveling. There may be a timid passenger in the way, but there may not. There is no system that will be full 24 hours per day, especially on the edges of the network, and optimal travel in these instances deserves consideration too. There is no rule against fast or slow lanes, passing lanes are anything else borrowed from the automotive transport world. One other point is that in tight spaces turning radii might be advantageously decreased, requiring a deceleration and acceleration. This all requires a much more robust and autonomous control philosophy than what is out there today because introduces a degree of chaos to the system, and managing that may be a challenge. Anyway, control is a very complex issue, and needs to be broken down into a number of posts, and this one’s purpose is to explore just one small aspect, that being not rear-ending a vehicle.
This discussion wouldn’t be complete without mentioning the California’s PATH program. In an interesting 1997 demonstration, a tightly packed convoy of 8 Buick LeSabres traveled under automatic control at highway speeds while maintaining close headways within 10 cm. This was done with autonomous vehicles communicating with each other and with proximity sensors. There was no central control.
Because of safety issues, there needs to be extreme redundancy in any system that controls the spacing between vehicles. One idea is to create graded zones that follow each vehicle like a comet’s tail, allowing a following vehicle to ascertain the distance to the lead vehicle. The idea may be visualized as in the picture below. Each following vehicle can have behaviors that are conditional upon these zones. The colors, of course, are just for the benefit of you, the reader. How the zones are created and detected is up for grabs.
The lead vehicle must create these zones and the following vehicle must be able to detect and react to whichever zone it is entering. While a proximity-measuring sensor would seem to do the same thing, I am interested in methods that are rely the simplest physics possible, to compliment the onboard sensors. These could be coupled with equally simple fail-safe systems. What if, for example, brakes were only NOT engaged when outside of the close range (red) zone? (Let’s assume no actual contact platooning for the moment) What about power? What if no power is available to vehicles in yellow through red? It is at least worth pondering.
To create such zones, there needs to be a way to influence something along the track that can transmit the information along to the upcoming vehicle. That information needs to change in amplitude or frequency as the lead vehicle gets further away. One thought is some sort of wave-guide. For example an acrylic rod can act as a fiber-optic transmitter. Notches sawn at intervals will light up brightly, but the effect will decrease with distance from the source. Measure the brightness at such a notch and you can tell the distance from the source. The same can be done with sound or radio waves.
Here is another very simple principle that ought to work.
This works on the principle of voltage drop through lengths of wire, although actual resisters, as shown, increase the effect. The down side is that I employ an actual electrified rail to transmit the voltage to the following vehicle. Anyway, it can be seen that voltage sent by vehicle A will incrementally decrease as it gets further away. Vehicle B can measure the voltage and know its distance from vehicle A. The diodes make the electricity only flow backwards in respect to the direction of traffic. The segmented rail would create metronomic breaks in the transmission, giving an accurate means to measure velocity.
I know. The idea of the electrified rail is impractical. Anything that is done to the track needs to be very inexpensive. Track based transmission, however, is immune to “line-of- sight” issues, so vehicles can “see” each other around curves. The other problem with all of these schemes as well as vehicle based signaling, (which would probably be the primary system, but that’s got to be a whole different post) is that a failure in the signal generation or pickup emulates the absence of a leading vehicle. A broken down vehicle must not become invisible. Ideally the whole thing needs to be reversed, so that the weaker the signal, the closer the proximity. That would be like the air brakes on trucks and trains. If there’s a failure, they engage, not the other way around.
10 comments:
This is a tricky issue, indeed.
I agree that ideally we wouldn't rely on line of sight. On the other hand, anything that would block line of sight should also be recognized as an obstacle?
If you're looking for something in the track that is more fail safe (that does not leave invisible vehicles), maybe something with capacitors that are gradually recharged, but discharged as a vehicle passes over the track? I'm not an engineer, so forgive my ignorance of this stuff. The closest I got was designing processors in a computer engineering course.
Dan VerHoeve responds - (using his actual name, in a shameless act of self-promotion)
Thanks for checking in, Alfransen. The capacitor idea is definitely worth pursuing, imho. I thought of the same thing but haven’t developed the idea, other than to recognize that it is a way to store information after the fact… That is to say it introduces the element of time as well as proximity, and uses simple current as the medium. I guess one way to have them reveal a dead bogie would be by allowing voltage to develop slowly around the bogie in question, so that a stopped bogie would build a greater voltage signature than would be the case if the bogie passed over an area quickly. This still has the problem of needing a cheap, continuous sensor though, or the conductive rail. One thought would be to try to use a long permanent magnet array on each bogie, and enough magnetic (Reed, hall effect) switches so that one would always be activated. Or there could be a continuous wound ferrous core, which would be like a linear generator, at least when bogies are moving…but I’m not sure about how to test for dead bogies with that one. At the moment I like my idea of plastic fiber optics, but now I’m thinking of a pair, with one lit and the other dark, with the bogie providing a close proximity reflective (transmission) means. That way the dark side only gets lit if a bogie is there, and the bogie doesn’t need any power to do this. Actually a quick web check reveals that monofilament plastic fiber-optic cable is readily available including “side-light” types, which are already designed to leak light. I wonder about using this in reverse for communications – transmit into the side of the cable and collect at the ends – It reminds me of the “leaky” coaxial cable technique (often used for sending radio transmissions into tunnels) promoted by Anderson’s PRT designs.
The best way to keep 2 people who can't see each other from colliding is to have a 3rd person keep track of their locations. If this 3rd person loses track of one, both travelers should be advised to slow way down until the missing one is found.
That's basically what a Zone Controller would do, in addition to adjusting vehicle speeds according to wind conditions and merges.
The PATH program proved that computers could use sensors to guide vehicles at highway speeds under controlled conditions. What it didn't prove was that drive-by-wire, without a backup system in case of failure (e.g., Toyota accelerator pedals), would be safe.
There's no inherent reason why a Zone Controller couldn't allow passing lanes (really just a long siding with a higher line speed), or different line speeds along portions of a Zone.
In any case, higher speed almost always translates into higher costs, which I think ought to be borne by those who value it highly.
Dan the Blogger is not a big fan of centralized or even regional control...
Hi, cmfseattle. As you undoubtedly expected, I disagree. If two people can’t “see” each other, they need walkie-talkies, WI-FI or mirrors or something. I think the introduction of a third party is a mistake. In my opinion, “wayside computers” and “zone controllers” are relics of a bygone era. I think the entire operations of both can just be included on-board every vehicle. That stuff was born in a day when vehicles were assumed to be deaf, dumb and blind. They required external, more centralized control. I also think that Irving and Anderson moved as far toward local control as the times would allow.
I mean, how big is this “zone? How many vehicles would it “control”? If it fails does the whole zone become unusable? Where are the zone controllers physically stored? How and where are they backed up? Is there any way a failure can cascade to other zones? It is hackable? The big question is, “What does it do for the vehicles that the vehicles can’t do for themselves?”
I am going to post more on this later this evening...
The natural size of zone is that between two merges / joins. This is like air traffic control - once the vehicle leaves the zone, it is handed over to the next one. A failing zone would not be taking in any traffic, so no new vehicles will come there.
how well do "walkie-talkies, WI-FI or mirrors or something" scale up to thousands of participants?
The big question is, “What does it do for the vehicles that the vehicles can’t do for themselves?”
they can communicate with other zone controllers via fiber optics.
This thread continues in the comments section of the next post.
Cabintaxi from the 70'ties used diodes in series at each meter to detect vehicle positions by measuring voltage. We won't need that type of thing today.
More interestingly they also used a system like the one you presented where a vehicle activates a mechanism which then gradually decreases in amplitude, thus indicating the time since the vehicle passed. I'm sad to say that I have forgotten exactly how this was done. I think it was some kind of a delay line (i.e. a cable with a lot of distributed capacitance and inductance) which was energized at a frequency using a coil. The delay line cable dampens the amplitude proportional to the distance (maybe squared).
I don't know how a vehicle could avoid destroying its input signal by its output. Maybe they used multiple frequencies or time slots. Today we could do such things at least.
The important trick they did was to short circuit the delay line cables periodically before merge points. Somehow they managed to get this to work so that a trailing vehicle could sense the _minimum_ of the distance to its own leading vehicle and the leading vehicle on the adjoining line, thus automatically handling the problem of who gets to go first through the merge. As usual the lack of hardware availability inspires smarter thinking! -- usually at the expense of higher cost and maintenance, though.
When it comes to control systems it seems that at least Vectus has an advanced control system such as you describe, where headway is speed dependent and each vehicle "drives as fast as it can". Well, in their case it would be a wayside decision (linear motor in track), but anyway.
Thanks Bengt, I did not know any of that. I am especially impressed that Vectus has such an advanced control system, since the obvious, easy way to control vehicles, where linear motors are in the track, would seem to be to activate the motors in equally spaced sequential “waves” that the vehicles would “surf” on. The “fast as you can” approach is SO much more complex.
It is also a good call to mention the maintenance implications of these systems. When I talk about a three-tier system or a consensus approach to control, I need to bear in mind the maintenance as well as the complexity of such an approach. Redundancy comes at a price. It would be nice to simply have all on-board computers talk to each other and cooperatively strategize in real time with an intelligent track. Unfortunately this leaves the city with a huge investment in an infrastructure that nobody, except the original vendor, can service. Hence my explorations in doing things the old way. This exploration is far from over, however, and “smart track” or a zone-controller might still be proven essential after all.
The first thought I have on this topic is that multiple systems are appropriate. To choose the right ones, note there are transmitters and detectors.
Some systems combine the transmitter and detector into one device (Radar), and others allow the transmitter and detector to be separated (the CabinTaxi signaling). Generally it's preferable to have both together because on the theory that you can detect that one or both components of the unit have failed after one false response as detected by another two working transmitter-detector pairs. But if the two are separate and a transmitter fails, then no all detectors that interact with that transmitter will produce the same false response.
One trick is that no system is without external transmitters, it's only a question of degree. For radar, the metal shell of the vehicle is a transmitter, although a fairly foolproof one as long as it hasn't been coated with radar absorbing material.
There's another way of classifying the sensors. Some detect the relative position of a vehicle and a fixed point on the track. Others detect the relative position between two vehicles.
If you're using a sensor that detects the relativity of track and vehicle, there isn't a control benefit to placing it on the vehicle. A vehicle can make no autonomous decisions using this data alone, it is dependent upon the data communicated to it about either where it should be, or where other vehicles are in order to make any non-default decision.
On the other hand, sensors that detect the relativity of two vehicles can allow control decisions that are autonomous. However, if that is the only source of information, then the vehicle must be able to detect the relativity of many vehicles to operate in an efficient and safe fashion. If it is only capable of detecting the relative position of a single leading vehicle, then a group of vehicles will be subject to a compression effect on stopping that will require instantaneous reaction time, or for each vehicle to have a higher maximum braking force than the vehicle in front of it, or a greater headway.
Also, I'd remind you that it's not a red-yellow-green, it's more of a yellow-red-yellow-green. At a close enough distance a slow down doesn't cause a higher speed collision, thus the theory behind coupling. It seems a trivial point, but it has a lot to do with control when you start thinking about failure cases.
You may have read it, but like many others here, I'd point you to some of Anderson's work, Fundamentals of Personal Rapid Transit, chapters 4 and 6 being the most relevant.
This text is very old, and even Anderson has different views in many areas, but there's also quite a bit of relevant material left out of anything more recent.
The most important part of this is the various considerations of how a vehicle would behave when all communication failed, and how the other vehicles would behave after all communication with a vehicle failed.
Through acknowledgement communications these two cases should always occur near the same time, but there will always be the possibility of a difference equal to the period of one communications cycle, or whatever is set as a timeout value for retries.
Post a Comment