Sunday, August 8, 2010

98> SPEED UP! SLOW DOWN!

Every now and then an alert reader asks a question so profound that I just can’t keep it buried in the comments section. This time the honor goes to cmfseattle for this question, posted under Control Issues Part II.

“Why would vehicles be going at different speeds, anyway? For any given segment of track, there is a maximum safe speed. Why not just have the vehicles go at that speed?”

The short answer is passenger comfort. A system like the one I envision is capable of much higher speed maneuvers than most passengers would be willing to endure. But let me put this in context. I do not believe that previous PRT designers have chosen centralized or zone-based control because it is better, but rather because they had to. Autonomy requires cheap, powerful computers, sensors and very fast and flexible communications. Since all of this is pretty standard stuff these days, these obstacles have been removed. Another factor is the nature of the network. Most PRT designers and vendors have envisioned systems where short trips are taken around a downtown area by hoards of passengers. Since they look at their systems primarily as a business venture, if a track segment wouldn’t be packed with vehicles, that segment would not be built. The strategy of only “picking the low hanging fruit” and then moving on to different city makes perfect business sense, and requires only relatively slow vehicles moving in unison. Let me give an example of how this philosophy creates limitations.

In most PRT systems, with the vehicle riding on top of a guideway or rail, the need for going fast around sharp curves is addressed by banking that curve. In the case of “y” interchanges, however, banking is impossible. This leaves these systems with two choices. One is to go slow, the other is to make the “Y” structure very gradual, creating the need for much more track. It has been pointed out that with frequent stations, a large percentage of the overhead track would, in fact, be double, creating extra cost and an unwanted canopy effect. The option of slowing for a curve is not needed because the vehicle is going slow in the first place. I do question just how smooth these interchanges can be made without slowing down, though. Going at a fixed, slow speed has other benefits. All curves can be sharper and banked at the optimum angle for line speed. The vehicles need not be particularly crash-worthy. Of course less expensive propulsion systems can be used as well. Finally, since they (PRT vendor/designers) see themselves having absolute control over all aspects of the system, there is no need to consolidate control. Some control could be with the stations, some with the vehicles, some with the track and some with a central system. This hairball approach is perfectly logical if nobody outside of your company will ever have to work on it. Anyway, within the context described above, a fixed speed which is controlled from outside of the vehicle makes perfect sense.

As soon as you move to higher speeds, however, the fixed speed approach gets much more cumbersome. Even with a hanging, self-banking design, there are limits to what kind of G-forces passengers will tolerate. Variable speed allows PRT vehicles to do what ordinary cars do for any particularly sharp turn, which is to slow down. But the faster they were going, the more time is spent in a transition speed. I would note that even with the fixed speed systems, there are always transition speeds anyway, such as entering or leaving a station or creating a slot for a merging vehicle. Longer, smoother transitions generally mean a more comfortable and energy efficient ride, something left out of the “line-speed” paradigm.

There is a trade-off that exists between speed and system efficiency and the nausea prone stomachs of a small but significant portion of the ridership. In a heavy traffic situation, there is only one choice. Slow everybody down to a point where the ride is acceptable to the vast majority and just let the rest be uncomfortable. With variable speed, however, if there are no motion sickness prone passengers on a track segment, everyone could go faster. The information for this could easily be gathered at the point of payment or stored in the form of account information.

Another advantage to variable speed has to do with inertia. Heavily loaded vehicles will tend to take longer to accelerate and decelerate, and should have longer headway distances. This might influence the how the vehicle behaves in merges and turns. These effects also become more profound with greater speeds. For example, why waste the energy quickly accelerating a 2 AM delivery? Even with regenerative braking, it would still be advantageous to coast to a stop if time and traffic allows. Everyone who drives knows that the acceleration/deceleration profile that is best for energy efficiency is not the same as for getting somewhere in a hurry.

Finally, it would be more efficient if the PRT behaviors included a bit of opportunism. Ever get on a freeway entrance ramp and realize that if you hurry you can slip into a spot without making anyone squeeze you in? In a PRT system, there could easily be very limited merging opportunities onto a busy track, and since relative positions would be known well in advance, the merging vehicle could “step on it” for a very long time to catch such a fleeting slot. This is essentially the “step forward” maneuver that Anderson envisioned but with many steps forward by a single vehicle rather than many vehicles all stepping back one step. This maneuver, again, involves straying from line-speed for an extended period of time.

In the end I think it will just be simpler, more comfortable, cheaper, easier to manage, and more energy efficient to program the PRT vehicles with behaviors that take into account the same sort of factors that influence drivers every day. How is traffic? Who (if anyone) is my passenger? Will I hold everyone up if I drive slowly? Will I benefit merging traffic if I speed up? Should I drive more conservatively because my vehicle is loaded down? Is this an emergency? Can I get through downtown before rush hour starts?

In the end, it all comes down to the PRT equivalent of intelligently using the gas and brake pedals. Any system that fails to address acceleration/deceleration properly is going to be uncomfortable to the rider and will constrain the options available to the track designer. The other option is to just go slow.

11 comments:

cmfseattle said...

Control Systems for PRT

Fundamentals of PRT, by Dr. Jack Irving, et al. (Chapter 4, Control Alternatives)

Control of PRT, by Dr. J.E. Anderson

Dan said...

Thanks cmf... Here is one more.
comment on the Anderson paper

Ryan Baker said...

Dan, step back a moment and consider all your assumptions. It's clear that some of them come from your experiences with driving a car. While that is a good source to draw from some are invalid.

One in particular is the assumption that acceleration will get you there quicker. In a PRT system, where there are no stop signs or stop lights that's basically invalid. Acceleration will influence a very small timespan of your trip.

A second argument you have for variable acceleration profiles is the efficiency factor.. i.e. more efficiently utilizing open slots. But as you've pointed out before, it's only the downtown systems that are likely to be so heavily utilized that slots are not readily available. In those systems the costs of adding more powerful propulsion that are only utilized sometimes is probably a greater cost. I'd think in those systems you'd have a default speed, and then if necessary a slower speed, but you seem to envision normal and faster.

Is part of the motivation here a personal desire to have a PRT experience similar to a rollercoaster? That's a perfectly reasonable personal desire, but I have a feeling that when it comes down to dollars and cents you'll continue to have to go to Six Flags for that.

There is a little bit of straw-man in some of your later arguments. You know that "fixed" speed really means a single target speed, not literally an inability of vehicles to use acceleration and deceleration for the purposes of merging, etc. Opportunism is then of course going to be utilized within the energy-capacity-safety constraints that everyone agrees too. Since as I've already pointed out, opportunism only benefits high capacity systems, not personal travel time, I can't see why anyone's preferences will differ.

This leaves us with the last case, that of vehicles that need a slower than normal speed either because of lading, some greater than normal phobia of speed, or some physical condition. Such things should be relatively rare I think. In a high capacity system you would do everything you could to minimize them.

Excepting the conditions of lots of stop signs and lights, the main way you get somewhere faster is a higher average speed, not acceleration. I used to drive, and actually, I did drive fast, but I didn't tailgate or any of that stuff because it was totally unnecessary. You either had a open lane or you didn't and all that darting about was generally for naught.

Why did some people drive slower than me? Well one reason was the law, not a problem with PRT. Another was safety, again not a problem. The final reason would be efficiency, which is still relevant, and yes, in lower capacity, longer distance systems I can see validity in supporting different speeds here (freight/passengers), but I'm going to suggest again that the best way of doing this is to have multiple tracks with different speeds rather than one track with multiple speeds (to clarify a bit.. my meaning is the multiple tracks aren't dual track (side-by-side over the same street), but rather two different parallel corridors, but that will depend on circumstances).

40mph is the speed generally bandied about as the best for efficiency, both within high capacity networks and energy wise, so I can't see any reason that won't be the slowest line you'd build, and that's still faster than almost anything except a highway. You want some PRT highways operating at 60-80-100mph right? That can happen, but it is probably a bad idea to make every track this way, just like it would be madness to set the speed limit of every street at 65mph.

Dan said...

I just want to reiterate a couple of points. First, the piece is about the context of universal control architecture. Do we base it on a fixed speed with some programmed maneuvers or completely variable speed? It is a fork in the road, design-wise.

There is an assumption, in the visions promoted by PRT companies, of continuous full capacity. This can only happen for extended periods if the route is turning away people at peak times. In a system that is sized for the traveler, not the PRT company, there will be more maneuvering room than is generally portrayed. There are a lot of off-peak hours in the day.

I think you should reconsider that unbanked turn (enter/exit) scenario, which will occur again and again especially downtown as well as the sharpness of turns that are possible without crossing over corner properties. These are 20 mph maneuvers, not 40, even less for the queasy. The roller coaster ride is an inevitable result of trying to build within easements. Laying out track for a reasonable, steady speed (even 30-35) could be VERY expensive, considering the cost of property. I have not played around with the optimal length of forks but I wouldn’t be surprised if they were over a 100ft. for, say 35mph. I fear the net result would be no access to neighborhoods where a good, smooth route cannot be found. You are not the first person to think that I have a thing for fast thrill rides. Actually it is the fact that high-speed vehicles doing slow maneuvers give the most un-roller-coaster-like ride. With track guided wheels and a minimal suspension, even ordinary city turning radii would be a thrill ride at anything over 25 mph. It would be nice if any open-source PRT OS would also be applicable to less g-force adaptable designs, (such as track-on-the-bottom)

A principle way you can cut travel time in PRT is getting ahead of merging traffic, an opportunity which may present itself again and again. This has a system-wide advantage because it clears the track segments behind you for other traffic. BTW, I envision the station interchanges as having a dual purpose of being a temporary slow lane to allow passing. (passing is not for roller coaster loving speed demons, but to allow people access to the system who would be most comfortable going VERY slow… not just motion-sick prone, but also the elderly, or to extend freight services to times other than 2 AM.)

Btw, if 40 mph PRT came to Houston, everyone would still drive. It’s just not fast enough for the relative distances involved. Speaking of driving, back to your first statement about drawing from the experience of doing it… I would just note that most of city driving is spent either accelerating or decelerating, smoothing out the transition between situations that allow more or less speed. In PRT one of those situations would be sharp turns or the lack of them. If we accept the use of low speed turn-offs (and I think we have to) then following vehicles will want to hang back, (like you do when a driver puts on his turn signal) and later move forward into the now vacant position. Another situation is slowing to allow a vehicle in. If that merge is done (for the comfort of merging vehicle’s sake) at less than the main “line speed”, then considerable velocity and headway adjustment will be required by everyone involved. If you smooth these maneuvers out for efficiency and comfort, then when, exactly, do you really get to go at “line-speed” anyway?

cmfseattle said...

In the shweeb design, the passengers have a lot of control over speed and switching.

Ryan Baker said...

You could use station interchanges for the purpose of vehicles that run slow in order to reduce drag, but it would be entirely ineffective for the purpose of passenger comfort. From a comfort perspective speed is irrelevant in linear sections, only acceleration and jerk is. Having multiple speeds on a line is going to cause more merging which will require more acceleration/deceleration.

Turns are the only are in which there is any significant clash between the goals of passenger comfort and your own vehicle's speed. Comfort in relation to merges/exits from/to stations are dependent not on your target velocity, but the velocity and proximity of other vehicles on the line, and thus would not be reactive to the riders preference but at the mercy of other's preferences.

Maneuvers are part of any system. Moving forward to make a spot, moving backward to open one, catching a train, breaking a train; I'm not aware of any modern control system that doesn't include those. But none of those give reason to attempt to run two or more different target velocities on the same track, and neither does comfort.

In terms of comfort, since the turns are the main issue, I'd suggest that the most effective way to make a trade-off for the preference of low tolerance to g-forces and time would be to avoid the sharper turns at the cost of a slightly longer route. But once again here, I think the best configuration is going to be one default with an optional "slow".

"Actually it is the fact that high-speed vehicles doing slow maneuvers give the most un-roller-coaster-like ride."

If this discussion is about suspended vs. supported design then I've misunderstood you. I don't think you need to be an advocate for multiple target velocities per track to support suspended designs.

Dan said...

Ryan, in regard to your first two paragraphs, all I said is there can be low traffic instances where variable speed can aid passenger comfort. What you say is not always true, but is true most of time. In low traffic, the merges ARE turns that can be taken as slowly as need be.

The point of my piece is that I argue the case for variable, adaptive speed. There is an inherent trade-off between 1) minimal cost/convenient, dense routing 2) passenger comfort. 3) speed.

With adaptive, variable speed control, every time there is sufficient space between vehicles, there is the possibility of managing those trade-offs intelligently, even based on customer preferences. If there are no spaces between vehicles then there is little that can be done. You may think I am getting ahead of myself with customer preferences but hey, that’s what us futurist, inventor types do…

With adaptive, variable speed control the inherent trade-offs mentioned above may be mitigated somewhat by allowing slowing for turns and merges. In exits off of long straight runs especially, slowing down would seem an appropriate. This would keep the interchange size smaller. Obviously this solution can be seen on every freeway.

A second point of my piece is that if you carry out the necessary maneuvers (you mention them in your third paragraph) smoothly, it represents a lot of time outside of any target speed anyway. If you slow for any turns or interchanges the time going at “line speed” is less still, compared to the time spent in an acceleration/deceleration maneuvers.

That is my basic argument for adaptive, variable speed, in a nutshell. We can’t really debate every instance here. Trying to balance the three conflicting objectives above cannot be done generically or perfected anytime soon. (and there’s more than three anyway… like energy use, mechanical wear and tear…) It will always be application specific, and never be ideal.

Make no mistake. Existing systems will jerk you around, or go slow, or take up a lot of space. (Or all three) Variable speed is not a cure-all, but I really can’t see any compelling arguments for fixed speed. Why “bake in” inflexibility if you don’t have to?

qt said...

"Why “bake in” inflexibility if you don’t have to?"

Sounds like it's time for someone to do another lecture on your design philosophy. I did it last time, so I'll leave it as an exercise for another reader.

Ryan Baker said...

"Why “bake in” inflexibility if you don’t have to?"

The obvious answer to that question would be because you have to. Vehicles, track, etc. have to have a maximum speed. Not having a maximum implies either infinite costs, infinite inefficiency, or truly both.

So, what is the maximum? Yes, you could have the maximum be slightly higher than the "default" speed, but what is the value in that? Relatively little I say since if the majority of vehicles top out at the default few vehicles will ever reach the maximum for an extended period.

A more efficient use of resources to allow different target speeds is separate tracks since multiple tracks expands capacity and allows both tracks to reach their target speeds the majority of the time.

I get the feeling you think there is more inflexibility involved than this intrinsic limitation. But I don't know what that is. I have a general inkling that you imagine that control systems that don't involve autonomy have some limitation.

But if they do, I don't see it. They can slow down for turns. In fact seems to me that coordinated operations allow for this better than full autonomy. They run a "open" speed and a "full" speed (i.e. 60mph and 40mph).

In past articles we've both talked about ways to deal with different speed track segments efficiently. The idea I liked the best was having separate "pushers" that contain the engine. This avoids having the costs of heavier motors when running on lower speed track, without requiring passenger exits.

Somewhat I get the feeling we are both advocating for the same thing with different words. The only real point of contention in my mind is you feel systems need complete autonomy, and I think that is the unnecessary limitation. I think we both agree that at least in the limited sense of some capability for autonomy (I see this necessary for failure scenarios only though), can be fairly cheap in terms of components. But full autonomy harms the ability for coordinated action, and that harms operations.

Dan said...

Thanks, QT!
Ryan,
I’m not going to go much further on this until I put out the post (on control) that I was intending to post last Sunday. I will say, though, that at least now I understand that in your mind, autonomy means less coordination, and you rightfully have problems with that. I believe autonomy enables more coordination, not less, plus enables an unbreakable system, which a the primary reason for autonomy in the first place.

John Cleeland said...

What does the track do and what does the vehicle do?
Beamways has flirted with roundabouts that in some locations require slow speed of 30kph for through traffic. Mister proposes grade separation that allows full speed. To turn on Mister, you need to slow down with the equivalent of an off-ramp. In road design, the exit ramp alignment has the first curve at high speed and the next curve (after suitable deceleration distance) can then be 10-15kph. Some routes will make sense only at high speed and some, more local, at low speed.
Can you get high speed options in the vehicle fleet? If so, the control system must allow for higher speeds for only some vehicles, subject to other traffic. Like driving your V8 amongst the others. On a single lane route, it won't get too far. Do we have minimum power on express routes?