This is a response to a comment from an Mission Local article covering the BART candidate forum. 

Hi Dimitri,

If we’re serious about providing mass transit for most people in the Bay Area, rail is not an option. The systemic problems with rail make it too expensive and inflexible to serve the majority of the Bay population.

You can see the symptoms everywhere in BART.

1. BART is proposing a $10+ billion dollar project over 20 years that will result in nearly the exact same service. It will have the same crowding levels and nearly the same schedule and energy efficiency.
   
2. Expanding the system today costs more than it did during original construction. ~$700 million per new station. After 44 years BART stations still are only within walking distance of ~3% of the population and 50% of BART riders drive to their station. This helps explain why the Bay Area has more cars per capita than Los Angeles.
   
3. There's only one vender of brakes for the old and new BART trains. This means the vendor can ask almost any price, and BART has to pay them.
   
4. New BART trains cost $2.2 million each and have 43 seats. For the same price you could buy 90 minivans for a total of 630 seats.
   
5. Trains STILL require an operator despite the fact that the system is essentially a horizontal elevator.
   
6. Replacing tracks cost 10x more than replacing roads.
   
7. BART rider fares don’t cover even the cost of operation, let alone maintenance and expansion.
   

These problems with rail are not unique to the BART system. They stem from the systemic properties of rail. The big ones are.

1. Centralized control. Trains don’t have steering wheels so they must be controlled central computer system that moves the tracks. This computer system is usually provided by a single vendor. The control system must be updated for every new extension or operational change.
   
2. Small volume parts. There are orders of magnitudes fewer trains produced than cars each year. This means that trains are many times more expensive to build than cars.
   
3. Incompatible infrastructure. Trains only work on their own tracks so to expand, they must build their own custom infrastructure.
   

To draw a parallel to the computer industry, rails are essentially hard wired. Before software, the logic of computers had to physically wired. To change even the smallest part of the logic, nearly the whole computer had to be redesigned. This was expensive and slow. Then software was invented so that changing the computer logic only required typing. Today, massively complex math can be done with only a few lines of code.

To change a rail system, you must physically rewire it by laying down new tracks.

I agree that some things (like calculus) invented before dial up internet don’t need to be improved. However, I’d argue that even spoons, forks and plates are improved by burritos and other increasingly popular utensil-less foods.

Knowing that any significant change to BART will take 10+ years, we need to plan for the technological advances over that time. We don’t want to have a ribbon cutting ceremony with trains that are already decades out of date. I don’t know if self driving electric BUSES will be the best fit for mass transit, but it certainly seems headed in that direction.

Thanks to limited public and private funding, there have been major advances in self driving capabilities and electric car performance. Now, private companies, including most major car companies, are investing billions of dollars every year to develop both self driving electric car technology. We can only expect the innovation in these fields to increase.

A self driving electric bus system also solves many of the systemic problems that plague rail.

1. Decentralized: Adding new busses, stations or roads don’t require integrating with a central system.
   
2. Interoperability: Buses can drive on any road so it would be easy to expand, and integrate other systems
   
3. Mass produced. The buses would be affordable because they are small (think minivan on a tesla chassis), produced by major car companies, and used across the globe.
   
4. High frequency. Small vehicles would enable lower wait times for customers by coming more often.
   

So if we’re serious about providing mass transit for MOST of the bay area, we need to start thinking in a new paradigm. It’s obvious from the “Better BART” plan its associated bond that BART leadership is still thinking under the old paradigm.

Personally, I think BART has the opportunity to create the world's best transit system in 20 years. This system would be one that where the majority of people the Bay Area to not need a car to access. World leaders would come ride it to get ideas for their own countries. This is not impossible, but would require vision and political will.  

Regardless, BART riders and taxpayers, should hold BART to a higher expectations.

​

This example with the car are similar to the situation we face with BART. The times match when BART was constructed (1972).  The costs of buying and repairing the car are proportional to the original cost to construct BART ($9.3 Billion in 2016 dollars), the estimated cost to maintain BART (>$10 Billion over the next 10 years). Determining the cost of replacing BART with a better system will be the topic of the next post.

I hope this was helpful to get an intuitive sense of the decision the BART Board and voters get to make in November.  Should we continue to maintain our Lincoln or should we upgrade with the times. 

I do support the 2016 BART bond but we need directors who know how to build a transit system that won't need another bond in 10 years. 

References:

• Cost of Original BART.
• Cost of Capital Improvements for next 10 years.

Traffic jams are sometimes caused by too many people trying to drive on road that's too small. However some traffic jams happen even when their is plenty of road space. This post explores why this happens with today's cars and how these traffic jams will be fixed with fast self driving cars.

To demonstrate the concepts  I'll use a simple web based traffic model. You can play with this model to see what else causes traffic jams. 

First, the simulation shows traffic on a circular road. At first the cars drive without issue, but after some time you can see speed perturbations (acceleration & deceleration) form because each driver can slow down faster then they can speed up. Eventually a jam forms where cars are nearly stopped which causes the cars behind them to stop. 

How can we fix this? There is plenty of room on the road.

You can completely eliminate these types of traffic jams simply by increasing the max acceleration rate of the vehicles to match their comfortable deceleration rate. This essentially helps pull cars out of traffic jams so that they don't slow down the cars behind them.

This is really hard in practice on public highways because no sane regulator will advocate for people to drive faster to reduce traffic jams, even if on paper it will be safer. Also, some vehicles like trucks simply can't accelerate fast enough to avoid causing traffic jams. There's no hope to solve this problem on public roads where there is a wide range of vehicle performance.

However, on roads where the specifications of all cars can be defined (like the simulation) these traffic trams could be engineered away by only allowing cars to accelerate as quickly as they decelerate. Who knows how much energy this would save, but its a lot. This is one of many reason why I'm advocating to create a purely self driving right of way on tracks BART tracks.  

All BART station graphs.

Here's the data source on BART's site: http://64.111.127.166/origin-destination/ 
Here's the ipython notebook used to create the graphs.

AutoBART is based on the assumption that autos can safely navigate the route quickly, drive close together and pull out of the right of way to pick up passengers. Here are a couple ways to test these assumptions.  

2d Simulation 
There are many 2d game library's that could be used to test the control models of autos. This would be helpful to learn the dynamics of platooning and the tolerances of staying within lanes. 

3d Simulation
Researchers are already using 3d game engines like Unity to simulate environments for self driving car software. To simulate driving in BART, CivilMaps could make a 3D map by mounting a LIDAR sensor to the front of an existing BART car.  Comma.ai released an app (chffr) to help learn from driving conditions. Maybe this could be used to show how hard/easy it would be to drive in BART (instead of on the road).  

1/10th scale cars on a setup track 
The self driving team at UC Berkeley has made an opensource self driving car (BARC) out of a 1/10th size remote control car. The parts cost $500 dollars looks like an involved setup. These could be tested or raced on a track in a warehouse.

Here is a video of one of the BARC projects.

List of students and researchers at UC Berkeley: ​http://www.mpc.berkeley.edu/people

BART is based on old technology. Here is an article presenting some of the data. 
​http://www.mercurynews.com/ci_29687067/has-barts-cutting-edge-1972-technology-design-come.html

Walnut Creek station had a fire on the third rail. ​http://www.sfgate.com/bayarea/article/Walnut-Creek-BART-Station-closed-by-equipment-fire-8759514.php www.sfgate.com/bayarea/article/Walnut-Creek-BART-Station-closed-by-equipment-fire-8759514.php

I've started a spreadsheet model of the AutoBart to calculate the capacities, speeds and dimensions needed to reach our service level targets.