« melbourne: a frequent network map | Main | email of the week: dept. of blindingly obvious ideas »

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d83454714d69e20147e045baba970b

Listed below are links to weblogs that reference basics: finding your pulse:

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

anonymouse

I could think of a nice way to do this illustration in color actually: have minutes past the hour at a given point along the line be a cycle of colors around the color wheel. This seems like a good experiment to perform on a network with lots of hourly or half-hourly routes actually, local examples of which include Santa Cruz Metro and Monterey-Salinas Transit.

David Marcus

I find two big challenges with my local pulse-based system.

One is the requirement it imposes that every route take about 35 minutes (we run 40 minute pulses). This means, for some routes, instead of running legible, direct routes down arterials, routes make semi-random detours into neighborhoods, apparently to burn time. This of course has a negative impact on the usefulness of the lines.

A further limitation is on our through-running routes. Forcing them to stop mid-route at the pulse point can add delays of as much as 8 minutes to routes that are only 30 minutes, end-to-end.

While pulses do seem to make sense for infrequent routes, for those of our routes running at 20 minute frequencies, I wonder if they wouldn't do better independent from the pulse system...perhaps gaining the extra operational efficiency to support 15 minute frequencies.

Brent

These are common in some of the lower-ridership transit agencies in suburban Toronto where a large proportion of their riders are connecting to/from GO trains (the commuter train system). Many of these systems will run "train meet" or timed transfer schedules during the morning and evening commute.

In some cases (e.g., Oakville) a good portion of the system is radial not around a downtown, but around a GO station. That simplifies the pulse network greatly -- the only scheduled connection that is really important is the one at the GO station (the network is set up radially to get most of the town to the GO station, so there is not much transferring), and there is less need to adhere to a rigid, even headway (in some cases the headway can be wildly variable because it is instead designed to meet arrival times for individual trains).

A different example is Guelph (west of Toronto), where there are several (a dozen?) radially designed routes that all converge on a single downtown transfer point at the same time every 30 minutes. Again, being largely a radial system, this is an easy timed transfer, but the challenge is that a combination of route extensions and congestion have made it more difficult to maintain the same round trip time -- a couple of years ago this forced them to reduce the headway to 40 minutes, which is not only much worse from a rider's perspective, but also makes schedule memorization more difficult (no longer "all buses depart downtown at the top and bottom of the hour").

AmandaKennedy

Why a pulse system hurts riders in my town of Stamford, CT: It reduces what could be a frequent transit network within the downtown core to an infrequent suburban bus system. Many of our routes merge as they enter the downtown, with 3 or more routes serving the same downtown-to-train station trip. Under the pulse system, I will see a stream of buses go by that could bring people to the station or back, but then it can be 20 minutes or longer before the next stream appears. Staggering the routes would mean that we could have a bus every 3 minutes at peak periods, and people could actually use the bus system within the core without regard to scheduling.
I'm not convinced that enough people transfer from bus to bus to make it worth losing out on frequency improvements made possible by these overlapping routes.

LX...

Totally agree with the benefits of pulse time networks.

However a particular challenge not mentioned is that it dictates a requirement for large bus interchanges capable of accomodating every bus on every route at the same time.

In a small hub with just a few routes this usually isn't an issue and can even be handled on-street.

In a medium sized city with say 30 routes providing bays for all 30 routes can be a real challenge in a town centre environment.

This scenario dictated a move away from pulse timing in the late 90's in Christchurch NZ even though most routes only run half hourly frequencies.

More recently Palmerston North NZ a smaller city with about 10 routes has also been forced when adding new routes to offset them from the main pulse to deal with space contraints in the city centre terminal.

It can be hard to justify millions of dollars for prime central city land for a facility that is empty for most of the time between pulses, versus say a much more compact facility that recieves more constant use, and the saving possibly put into say better service levels.

Difficult trade offs as ever.

francis

Here in the Bay Area, since we have so many transit agencies, the pulse scheduling gets pretty awful, especially since each changes their schedule at different times of year based on their fiscal cycles. So there often is the every 20 minute bus meeting the every 15 minute train.

However, on the late night (after midnight) system of routes, pulses are used, as service is hourly and streets are empty.

David M

I once asked Victoria Transit (BC Transit) why they didn't use time transfer clock-face operations (ie pulse scheduling) and I was told that their interest was to maximize service hours. The problem as they saw it is that pulse scheduling requires all buses arriving and departing to be on schedule, so you need to bold redundancy (recovery) at the pulse points. The bus needs to arrive and maybe wait for 5 minutes before departure. I was told that this could eat up 20-30% of the operational hours. Because it is not tied to a timed-connection, or pulse operation, Victoria is free to run buses at whatever frequency best meets the demands of the route. Therefore, you'll see buses running every 8 minutes, every 11 minutes, every 23 minutes, or not an approximate frequency where the times between buses varies a few minutes between every trip. In a small system like Victoria, instead the routes run on schedules that do not adhere to the clock-face. Connections are made, but not every trip connects exactly - the schedules provide connection information, but for the most part, there's a need to look closely at the schedule for each trip.

However, this systems works really well at Ladner Exchange south of Vancouver. Ladner is a major regional exchange with buses converging from South Surrey, South Delta, Tsawwassen ferry terminal, Vancouver, North Delta and Surrey as well as local routes. Bus lanes help buses remain on schedule despite the distance, traffic and obstacles like the Deas Island Tunnel. It pulses every 15 minutes, with every 30 minutes being a major pulse when just about every service is present. Quite often, supervisors are present at Ladner and buses are held for late running connections to ensure the connection is made - or overload buses are provided to take the late traffic in some situations. The opening of Bridgeport Skytrain Station has reduced the importance of Ladner as a transfer point, but it is still a a major connection.

Other bus stations around the Vancouver region (outside of Vancouver) have similar setups. But Ladner is the best to observe this happening in the Vancouver region.

Edmonton exclusively uses this system. Every route in the city is a timed connection at a transit centre (bus exchange). The drawback is a through bus ends up waiting at an intermediate terminal for its timed departure, slowing the through trip down. Another downside is you are locked to the pulse for scheduling. If you have a basic 15 minute pulse (as Edmonton has) then you're limited to frequencies of 5, 7.5, 15, 30 or 60 minutes to provide the same connections every hour. 45 min is possible but the connection times would vary on the odd and even hour. 10 min or 20 min frequencies would then lead to connections only made on certain pulses, but not all. And odd frequencies like 11 minutes or 23 minutes would provide virtually no connections. If you review Edmonton schedules, you'll see they stick to the 7.5, 15, 30 60 frequencies, with a few odd 45 min frequency routes in low ridership areas. You won't find any 10 min frequencies or other odd frequencies, with the exception of the LRT. Personally, I think this creates a higher expense in providing the service and makes it harder to fit the frequency to the ridership demand - but it's a trade off for the betterment of the system.

David M

To add to my last post - if a route has enough demand and is running a frequency of say 10 minutes or better, I think it's best that that route be removed from the pulse and just run. The frequency will ensure reasonable connections, at least when transferring to the frequent bus. The other direction will require some planning on the part of the rider. But then, isn't this how metros operate? LRT, Skytrain, metro and subway trains don't hold at stations waiting for buses.

Rob F

One of the biggest challenges to pulse operation is, of course, that it relies on a level of reliability that is difficult (or not a priority) to maintain.

I've never been a scheduler (and haven't even played one on TV), but when I studied scheduling a couple of decades ago it seemed to me that a bigger challenge for timed transfer operation is the lack of scheduling tools to produce schedules that make good connections. Things may have changed more recently, but at the time, scheduling programs were focused entirely on optimizing labor allocation from the operator's point of view, and not focused much on optimizing trips times for customers making connections. Trip building is (or was, at least) done separately for individual routes or groups of interlined routes. I've had the impression that making timed meets work requires a lot of manual adjustment of schedules by the scheduler, and all of those must be maintained when travel times change on individual routes.

A scheduling system that focused instead on optimizing customer connections would need to include a network-based trip building module, which would shift trip start times in order to prioritize meets at designated locations. This is a classical optimization problem that the math programming experts who make scheduling software ought to be able to solve and automate if transit agencies wanted it done. Tools like this would make timed meets more achievable for routes with multiple pulses, but also for simple transfers between two high-volume routes at a bus stop. I'm way out of date on the state of the art for scheduling, and maybe new tools are available that make timed meets easier to schedule, but if that's the case I haven't heard of it.

Brent

To Amanda's point, Hamilton (Ontario) has a way of partially mitigating this while maintaining the "pulse" principle. Hamilton is characterized by an escarpment face running east-west along the entire city, so that there are a limited number of roads linking the upper city (suburbs) to the lower city (downtown). In the upper city there are a half-dozen or so parallel north-south routes at half-mile spacing that ultimately converge on one road descending the escarpment face to downtown. They run at common headways but are staggered so that, if the routes are numbered 1 through 6, the odd numbers leave downtown at the same time and the even numbers leave half a headway later.

anonymouse

I think with headways of 15 minutes or less, you don't really need to coordinate with pulses too much, just make sure that you don't have connections with only 1 minute (or generally inadequate time to transfer), because there nothing as frustrating as watching your train or bus pull away as you're running to catch it, and knowing that the schedule is designed to do this just adds insult to injury. With a system of 15-minutely routes, it's probably optimal just to have pairs of them arrive at a transfer point 7.5 minutes out of phase with each other, to minimize dwell time for through riders and still provide easy transfers between the two routes.

Drew Adamick

@ David M

Speaking as an Edmontonian using Transit, the pulse system here does have it's advantages (that is, you know that almost all of the buses at a transit centre are going to leave more or less at a certain time) and drawbacks (the dwell time at a transit centre- waiting for 5-10 mins at Jasper Place [which is located in a fairly seedy neighbourhood] on a bus while try to get to/from places like West Edmonton Mall].

However, my biggest gripe with the pulse system here is not the actual system itself, but rather the weird timing of the pulses. Take West Edmonton Mall for example. On weekends and late nights, the pulse is every :28 and :58 past the hour, however, almost all of the stores in the mall let their employees go at :00 and :30, which forces transit users to wait 25+ minutes at times waiting for the next bus, which could be a deal-breaker for transit riders. A :15 and :45 would be better imo. I've tried to bring this up several times with ETS but so far, I've heard little in the way of tweaking the schedules.

Alon Levy

Best industry practice for clockface schedules is in the German-speaking world, featuring both timed transfers and overlays that produce high frequencies on common trunk lines. The clockface schedule is rigid enough that some of the smaller systems just indicate how many minutes after the hour the trains leave. See explanation of the Swiss and German pulse system here, with the obligatory dig at some parts of the US for not implementing nice transfers.

Brian McCann

I was wondering who would mention the Swiss system. Correct, Alon; it was a revelation to me about the Swiss system that operating on a clock-face schedule dramatically increases the connections. There is little room for error, as the Swiss are noted for extreme punctuality. It seems to work.

Closer to home, Columbus has a modified pulse system, resulting in long bus "line-ups" waiting for about 5 minutes at the Statehouse. Unfortunately, what little retail around Capitol Square was pretty much killed off by the line-up.

ChrisK

I live in a city that's had a pulse system for 30 years. What kills me about pulses are four things: dwelling time, route backtracking, distortion of geography and really crummy areas to wait for the pulse.

The first two issues anger any passenger wanting to get where they want to go in a timely manner. Operating at 30 minute headways with a ten minute dwell time at transfer centers is bad enough. Forcing your passengers to wait in some huge flat concrete suburban facility with little shelter in a sea of parking with no access to urban goods and services is going to cause any prospective rider to think twice after a few trips.

Additionally, because you focus on these "centers" of activity your transit facility must be large enough to house the peak number of vehicles in a pulse, leaving the area vacant most of the time. To an urban design specialist this is terrible and is unlikely to bring about any form of transit oriented development. In our case there has been zero TOD built near our transfer centers in the last 30 years. The route backtracking issue irks me because it goes in the opposite direction of a grid system, which has been discussed so many times on this blog. We also have a great number of east-west lines, depending upon the transfer centers for the majority of their ridership, but little north-south connections to complete a grid. So in our case, the transfer centers are acting as a barrier to transitioning to a grid system.

Can someone give me an example of a good mid-sized pulse system in North America that has attracted transit oriented development?

AL

Following with the Edmonton comments above, Edmonton Transit has done a reasonably good job of setting up pulses at transit centres. However, what they do poorly is recognizing that high usage routes *between* the transit centres should not always operate strictly to the pulse. This is especially true if the transfer point is a major destination in its own right.

For example, between the University and West Edmonton Mall, most riders are forced to endure long meandering routes through neighbourhoods where only a few get on and off. While arriving at a pulse is convenient for riders wishing to make a connection, those wishing to go to a destination at the transit centre just want to get there quickly!

Daniel Howard

A good map of a pulsed network is offered by San Francisco MUNI for their night buses, which run hourly, with timed transfer points circled in yellow.

Jarrett at HumanTransit.org

Thanks, Daniel.  I've updated the post to include it.

Nicholas Barnard

Has anyone looked to the airlines for scheduling tools? It'd seem their older banked hub system is exceptionally similar to pulsed transit systems.

Andrew

Whatcom Transportation Authority in Bellingham has done a great job with a Pulse system with one big hub and one secondary hub. They had the fastest percentage growth of ridership in 2006-2008 and they also did a high freq. route network. Unfortunately they had to cut Sunday service recently due to their volatile sales tax funding source.

Mark

Having rode the pulse based system in Madison, WI for many years, I came to truly hate the system. When I was riding, Madison ran with 30 minute headway, and there were six buses that I could take to get to/from work. In theory I should have had an effective headway of 5 minutes (30/6), but the reality was that 6 buses would come in a clump over about ten minutes followed by no buses for twenty minutes. Combined with the fact that the peak hours buses last crossed downtown around 5:30, getting to and from work was a constant annoyance. (The UW campus' deliberate removal of parking lots to build buildings was a whole separate annoyance. Reducing supply to increase demand? Brilliant!)

Alon Levy

The huge suburban parking lots with no shelter have little to do with a pulse. They're a separate dimension of transit planning, one that can be taken care of regardless of transfer timing. In Switzerland, Germany, etc., they've learned this and make sure to place train stations within walking distance of as many people as possible, using better acceleration and short dwell times to maintain an acceptable average speed. They make sure all the stations have shelter, and avoid park-and-rides, on the theory that if people need to drive for part of the trip then they'll drive for the entire trip.

Wad

ChrisK asked Can someone give me an example of a good mid-sized pulse system in North America that has attracted transit oriented development?

The North San Diego County Transit District has done a phenomenal job of building excellent transit centers for what is crummy bus service (lots of 60 to 180 [!] minute headways).

The Sprinter DMU system that's in place now serves all of those transit centers, but those transit centers were built for buses exclusively for decades.

Oceanside is a multimodal transit center with bus, Sprinter, commuter rail, Amtrak and Greyhound -- as well as civilan access to Camp Pendleton! -- all just a football field's length away from the beach. Vista has a major shopping center. San Marcos built its transfer center at a community college. Escondido has its transit center on the edge of a small-city downtown.

Eugene Wong

Hi Jarrett.

Thank you so much answering my question. I kind of hoped that you would have condemned 20 minute frequency in general, but this is good, too. ;^D

I appreciate the information, because, if I make any suggestion to transit companies, then I can make a more informed suggestion. Also, it's always good to know what our transit company has to go through to get us from point A to B.

Based on what you said, I like the pulse system when there are not that many bus routes, but if there is a parallel bus route, then I would appreciate it if that bus route would run at a different time than the other bus route, in hopes of creating more frequency for through traffic.

@DavidM

I agree with you about Ladner, and 10 minute frequencies.

@anonymouse

I agree with you about 7.5 minute wait times. It just seems so much less stressful. It's not short wait, but it isn't a long wait, either. It is long enough, though, to almost guarantee the transfer.

--
Sincerely, and with thanks,
Eugene T.S. Wong

Pete (UK)

Interesting article. The Pulse system is unknown in the UK as far as I am aware, which is a shame. Although it is common for bus routes to be 'interworked'. This means each vehicle may be sheduled to operate more than one route during its duty, but this is for operational convenience, and is not known to the public. In my home town, Chippenham, Wiltshire, we have a town minibus service advertised as routes A, B, D and P. In reality these services are actually registered as route 44 for operational convenience.

http://www.faresaver.co.uk/timetables/4c8de3ff24ad7.pdf

http://www.faresaver.co.uk/index.php

We do have bus stations in most towns and cities , but these tend to be seen as termini rather than interchanges. This could be because UK bus fare systems do not allow for transfers, unless you buy all-day tickets.

This is the new bus station in the city of Bath and is typical of a UK bus station:
http://3.bp.blogspot.com/_af3xMs0RPlA/TLwpdzodwGI/AAAAAAAAAfA/xiB8qJ_28u0/s1600/00bathbusstation.jpg
How the public see it:
http://farm4.static.flickr.com/3533/3812755282_e8fdc0cf90.jpg

John
Note the complexity of the scheduling task implied by multiple timed connections along the same line; it's an impressive commitment.

This reminds me of the challenges in synchronizing a traffic signal network. You kind of have to start with the busiest intersection and work out in every direction to get the timing offsets right.

Joseph

Ste. Therese north of Montreal operates an excellent pulse system which also pulses with the commuter train. This has attracted big-time TOD.

Paul C, Vancouver

The problem I see with pulses is they slow down the average speed of a route to the average amongst all the routes.

Also what happens if a bus on one route is behind schedule should the bus on the other route wait.

It would seem that this would only work if at the pulse point every bus on ever route waited a good 5-10 minutes. This would make sure the connections are happening. But it goes back to my first point in that it lowers the average speed of each individual to the average amongst all the routes. If you lower the wait time at a pulse point you increase the average speed but now your run the chance of people missing their connection.

mikef0234

I think a utilitarian argument could be made for or against pulses based on minimizing the total wait/trip time for all passengers entering/exiting a potential pulse point. It might go like this:

If the total time wasted by all the people on buses with the pulse point is less than all the time wasted by people waiting for the next bus without the pulse point, then the pulse point makes sense.

Assume it takes a few minutes T_dwell extra for a bus to enter/dwell/exit a pulse point. Assume some fraction f_transferring of people transfer there. Assume all buses operate on some constant headway T_headway. If T_dwell > 1/2*T_headway*f_transferring, then the pulse wastes more time.

This calculation oversimplifies things but it meshes with what the comments are saying about timed transfers.

If the dwell time is 5 minutes, the headway is 15 minutes, and 2/3 of people transfer, people are waiting the same time on average with or without the pulse.

If the dwell time is 5 minutes and the headway is 10 minutes, the pulse wastes people time on average even if everyone transfers.

Pulses start to make sense once the frequency gets much worse than every 15 minutes. Between routes with regular, frequent service, pulsed transfers waste more time on average than uncoordinated transfers.

Andrew

In Kingston, Ontario, a modified system like this is used. During the daytime from Monday-Saturday all routes run every 30 minutes but there are several sections where routes overlap on schedules staggered 15 minutes apart (routes 1/4 and 2/6, 6/71 sort of, also 12/12A and 10/71 in rush hour). There are two separate "pulses" every 30 minutes at the five main transfer points, Downtown, Kingston Centre, Gardiners Town Centre and St. Lawrence College, and all route segments are timed so that they are 15 (or 30) minutes long. The trouble is, there are a few places where weird detours are made to make the route segments 15 minutes long, particularly on Gardiners Road where routes 6 and 71 both make two different long detours through mostly residential neighbourhoods because the route between Gardiners Town Centre and Cataraqui Centre is very short relative to all the other segments between "pulses" and could be driven in less than 10 minutes easily. This system works well in small systems where all routes are converging on one or a very small number of transfer points (or in a larger system, where several infrequent routes which connect to a more frequent service at one transfer point operate on a "pulse"), but it has trouble when applied to a network of routes connecting at many transfer points.

Tessa

I remember wondering why the heck the 240 bus in Vancouver would always stop as soon as it entered North Vancouver, until I realized it was timed to meet with the 246, which went up the hill to Highland Boulevard. The only problem with this is even late at night, this often meant waiting five minutes or so on a main line bus route, while the 246, which often arrived with just two people on it, almost always arrived afterwards and didn't wait at all. Always bothered me, even if the idea was quite helpful in theory.

Bob Davis

Historical note: One of the most famous "pulse" operations in the US (although I doubt if it was called that) was the Fort Collins (CO) Municipal Railway, which had three streetcar lines operated with Birney cars that finally closed in 1951. The cars would meet at the downtown junction on 20 or 30 minute headways, and then go off to their respective terminals. I never saw this in action, but several of the older trolley fans made the pilgrimage to see the last Birneys in revenue service in the US.

Edward Re

Great idea! I was visiting a sister in law in Wagga Wagga NSW - population ~45,000. She had a client who was going to take the bus home. Because of connections, it would have taken about an hour. So she gave them a lift home, which took about 10 minutes.
This got me thinking that even if you have a direct bus, with the 30 mins or whatever frequency, it makes them fairly impractical. Why not have a taxi/bus, like the services from the airport? The driver enters the destinations in the GPS (at Seattle this took him about 20 minutes - pretty $%^ annoying), and then does a travelling saleman path to drop everyone off.
In the optimum situation, a passenger SMS's the bus their location and destination. The central server should work out the travelling salesman paths for pickup/dropoff (you could even use Google maps, which works out routes in a fraction of a second - even for multiple destinations). It should be cheap as you need fewer buses to cover the area, and every service would be door to door like a taxi.
In a small town like this, having a pulse service would require multiple buses with few passengers.

Jarrett at HumanTransit.org

Edward.  Even in a small city, and for people who don't have the option of driving, people value the sensation of freedom.  I use a lot of taxis, and I always feel vulnerable to their essential randomness.  Small-city taxis do have a role, especially outside the main daytime travel hours and for people who can't use fixed routes, but I think that even hourly fixed routes that you can count on, and built around a post, offer more "spontaneity" than relying on taxis does.

Annonymous

The door-to-door airport shuttle works ok when each vehicle is only transporting 2-3 passengers. After that, travel times get extremely unpredictable. I experienced this first-hand taking Super Shuttle home from the airport. By car, the trip takes about 35 minutes. The first time I used them, I was the first to be dropped off, and the trip took 35 minutes after a 30 minute wait time. The second time, I was one of the last to be dropped off, and I was on the shuttle van for nearly 3 hours, extremely hungry and thirsty, with no ability to stop for food or water along the way (with regular buses, food/water stops are easy), again after a 30 minute wait.

After I got home, I did a comparison of all the options:
1) Taxi - $65, 35 minutes
2) Super Shuttle - $35, 1-3 hours (including wait time)
3) Fixed-route buses - $16.25, 1.5 hours, reliable trip.

The result: When I had to make the same trip again, I sometimes used regular buses, sometimes taxis (especially when arriving late at night), but I never again rode Super Shuttle.

The comments to this entry are closed.

the firm

Jarrett is now in ...

Related Posts Plugin for WordPress, Blogger...
Related Posts Plugin for WordPress, Blogger...