It’s time to upgrade the train control system. Let’s stop wasting money on obsolete technology and parts. Stop wasting riders’ time with delays. Improving light rail, and transit priority, will be difficult at best if we don’t upgrade our train communication system.
Riders are tired of being stranded by Muni’s light rail service. Why does it seem like service has gotten more unpredictable? What’s so difficult about getting trains to run on time? We’ve heard that the old LRV3 Breda trains are getting harder to maintain, and harder to find parts for. We’ve heard about mechanical issues with the new Siemens LRV4s — including doors, brakes, wheels, and couplings. We’ve heard about the limitations of transit signal priority, causing too many delays for trains stuck in car traffic.
But what about the Automatic Train Control System (ATCS)? Shouldn’t it be helping to speed trains through the subway? Why do we still get stalled between stations or when entering tunnels? Why is service so unreliable even when it’s underground? What is the ATCS, what does it do, and what role does it play in helping Muni run reliably — or not?
How it works
Muni has a dual train control system. From the era of manual control there are safety buffer zones, marked by signals, keeping trains separate along blocks of track. The ATCS improves on that system with moving blocks — the same buffer zone concept, but the buffer zone moves with the train. Instead of fixed signal lights, the ATCS defines buffer zones based on readings from an antenna on the train and sensors embedded in the tracks.
This system speeds things up considerably (and has successfully prevented train-to-train collisions, which is its primary purpose). Where manual control maxed out at 24 trains per hour in the subway, or one every 2.5 minutes, the ATCS can potentially deliver 45 trains per hour, or one every 1.3 minutes.
The ATCS is dependent on a digital “handshake” between a train and the controller, and there are specific locations where the handshake happens — upon entering the subway and at a handful of spots along the subway. There are several ways the handshake can go wrong. Wheels can slip in the rain or on an incline, causing a mis-read. An antenna on a train can be off. The sensors in the tracks can be faulty.
If a train doesn’t complete a successful handshake, it must run in manual mode. Instead of going up to 50 miles per hour, it maxes out at 25 miles per hour. The manual mode train also maintains more space ahead of it as dictated by those fixed blocks of track. A train can try for the handshake at successive sensors, but if none work, the train continues manually through its whole route before going back to the yard.
If you think delays, bunching, and gaps are getting worse, you’re right. Just in the last couple years, there’s been a marked increase in what Muni calls Non-Communicating Trains (NCTs) — those operating in manual mode. In 2016 there was a low of 16 occurrences a day. This past February, we hit a high of almost 40 NCTs a day.
Each failed handshake reverberates throughout the system. The NCT impacts every other train. A 15-minute timeout caused about 2½ hours of residual congestion. And, the failures happen on both older and newer trains, clearly demonstrating that the ATCS is the primary issue.
It’s getting old
The last major upgrade to Muni’s train control, which brought us the current ATCS, was done in 1998. That upgrade is infamous as the original Muni Meltdown.
It’s now over 20 years later. Parts for the ATCS are becoming increasingly rare, causing delays in repairs or replacement. The technology is increasingly dated — when the computers went down this past February, a person had to physically walk a 5¼” floppy disk into the server room to reboot it. That hour of lost communication caused another 2 hours of residual congestion.
Expertise is fading away. Engineers and staff familiar with the technology, nevermind expert on it, are increasingly retiring out of the business. The vendor’s standard reply to requests for repairs or upgrades is that it will take at least twelve months.
For example, new crossovers were put in the Twin Peaks Tunnel during last summer’s construction. However, they aren’t signalized yet because SFMTA wanted to re-open the tunnel as quickly as possible. Until the signals are ready, hopefully next summer, the crossovers can only be used in emergencies for breakdowns.
There are other issues with the Muni subway that compound the ATCS issues. Because crossovers aren’t signalized yet, turnarounds take too long. Crossovers can stop traffic for at least 4 minutes in each direction. Shuttle trains, which should alleviate rush hour crowding by staying in the tunnel, therefore have to leave the tunnel and spend time in traffic going out to St. Francis Circle to turn around.
Muni schedules more trains than the subway can actually handle, and ends up tripping over itself. As part of the current analysis of subway service, SFMTA is re-evaluating what the actual optimal capacity is, in order to design the system appropriately.
The Sunset Tunnel isn’t able to take full advantage of the ATCS because it isn’t secure enough. Due to the risk of intruders and concerns for safety, the N-Judah has to remain at or below 35 mph through the tunnel.
What about the surface?
Having six lines function independently on the surface before merging in the subway is a challenge. At other times we’ve discussed separating subway service from surface routes for better speed and reliability. For now, we’re going to consider better train control as the solution.
As things work currently, trains on the streets are basically manual and signals act independently. The train’s location is communicated by an on-board computer talking with specific immovable points embedded in the tracks.
If we had full signal preemption, we might know better when trains will arrive to the portals. Preemption means interrupting traffic signals along a train’s route so the train always has the green light. This practice is standard in many cities, but not in San Francisco. It’s considered too disruptive to cars, pedestrians, and cyclists. (SFMTA uses transit priority in some places by holding a green light a little longer or making a red light a little shorter, but not full preemption.)
So each train’s arrival into the subway is very unpredictable, and controlling each approach is too much to coordinate in real time with the current technology. This unpredictability causes backups at the portals (and inbound to Van Ness) as priority is practically determined on the fly and it takes a minute to switch the tracks accordingly. Obviously these back-ups affect the functioning of the whole subway, and the entire system.
Where do we go from here? Time to upgrade!
We could rebuild the system we have for around $110 million. There doesn’t seem to be any good argument for this. As we’ve noted, this system is becoming obsolete.
We could upgrade the ATCS to a current version for about $150 million. We would still have inflexible points of control. We would still be only controlling trains in the subway, and not so much on the surface.
Or, for $200 million, we could get a new Communications-Based Train Control (CBTC) system based on current technologies like wifi and cellular. Muni light rail would have real-time location communication instead of static check-in points. Software can do a better job of predicting faults and providing redundancy.
With CBTC there would be better coordination with existing traffic signals. Trains could have true priority and preemption where it makes sense. There would be better system-wide sequencing, to actually address bunches and gaps in real time. There would be better arrival predictions, allowing for smoother coordination as trains merge into the portal and onto the same tracks. Trains would be connected to the system before they leave the yard — no more waiting until a train gets to the portal to find out if the ATCS handshake will work.
The roll-out could be a lot smoother than in 1998 by putting CBTC equipment on trains that already run on the ATCS. Muni could continue using the existing system while testing the CBTC in service to collect information and program as needed without interfering with existing service.
In one scenario, the CBTC system could be implemented over the next few years. Since that’s before all the older Breda trains are scheduled to be retired, Muni could limit where the Bredas operate as they roll out the CBTC system on the new LRV4s.
It’s possible that by winter of 2020, SFMTA could have a request for proposals out to source a new CBTC system. Over spring and summer of 2020 they could continue incremental ATCS improvements, including finally having the West Portal crossover ready for use. Theoretically, the new CBTC could be rolling out in 2023.
You can help!
Now that you know all the details, tell your friends. Share this story. Mention the need for new technology. Build support for the coolest thing ever — not a flashy new station or subway, but real train control for faster, more reliable, and evenly-spaced train service! Support our work to help push this forward.
Of course, we don’t intend to wait four years to improve our train system. What can be done to bring relief to riders now? We need more pilot programs like the one limiting left turns at West Portal. We need to target places where trains shouldn’t be stuck in traffic, where we can put transit first now to serve today’s riders.
Yes, real robust train management will require a new system. But we’ll also be pushing for changes to the system we have today to build the momentum for trains, and riders, first.
Bring your tech skills to SFMTA!
The Transit Technology Group is hiring! They’re responsible for the scoping, management, and integration of technology systems that directly support transit operations — including the ATCS, the NextBus system, the Harris radio system, and more. Check out the latest job listing.