Musk says a Model X derived minibus has "surprisingly high" people density.[1] The standard Model X seats 7, so I'd say at least 10.
I'm assuming similar volume-stretching tricks as the Model 3, (pushing the dash forward, glass roof for headroom) in a microbus package.
> All forms of transit exist on the tradeoff surface between these possibilities. As population density goes up, it's easier to find lots of trips that are close enough in the five dimensional space of origin, destination, time.
I think the real value of "big data" (something Jarrett dismisses as "big detail" in the article) is that we can plot every trip in that 5D space, deriving a "typical" density curve for any given day of the week. With that, the fleet management software can optimally allocate vehicles.
Clustering nearby points (trips) in this 5D space should optimally combine trips to minimize detours to individual houses (by grouping in the origin/origin and destination/destination 4-planes), combine trips across town (by grouping in the origin/destination 4-plane), and dynamically adjust the wait time to combine trips to/from low traffic locations (by grouping in the time/origin and time/destination 3-planes). Really it should trade off between all of these factors.
The challenge, of course, is that you don't know everyone's route ahead of time. Good prediction is important.
Essentially you're dynamically adjusting the bus schedule and route in response to spatial and temporal demand patterns. And with 10 passengers and no aisle[2], the maximum passenger density is similar to a bus (the "geometry" problem) So it seems like this can be more efficient than even a fixed bus schedule.
It seems like a decently large (100k+) multi-agent whole city transit simulation could settle the matter one way or the other. Who's plugging autonomous taxi routing into these?
This[3][4] looks promising. Accounts for induced demand. More publications on his website.[5]
[2] "Traffic congestion would improve due to increased passenger areal density by eliminating the center aisle and putting seats where there are currently entryways, and matching acceleration and braking to other vehicles, thus avoiding the inertial impedance to smooth traffic flow of traditional heavy buses." https://www.tesla.com/blog/master-plan-part-deux
I'm reminded of how the Chaos Congress organizers persuaded the Leipzig public transportation agency to run around-the-clock service to the congress: He showed them the curve of devices logged into wifi for the last year. It has a steep decline somewhere at night, which is when people leave.
With all the location data Google & Co have they should try their hand at making plans for public transport systems (if they haven't already).
Musk says a Model X derived minibus has "surprisingly high" people density.[1] The standard Model X seats 7, so I'd say at least 10.
I'm assuming similar volume-stretching tricks as the Model 3, (pushing the dash forward, glass roof for headroom) in a microbus package.
> All forms of transit exist on the tradeoff surface between these possibilities. As population density goes up, it's easier to find lots of trips that are close enough in the five dimensional space of origin, destination, time.
I think the real value of "big data" (something Jarrett dismisses as "big detail" in the article) is that we can plot every trip in that 5D space, deriving a "typical" density curve for any given day of the week. With that, the fleet management software can optimally allocate vehicles.
Clustering nearby points (trips) in this 5D space should optimally combine trips to minimize detours to individual houses (by grouping in the origin/origin and destination/destination 4-planes), combine trips across town (by grouping in the origin/destination 4-plane), and dynamically adjust the wait time to combine trips to/from low traffic locations (by grouping in the time/origin and time/destination 3-planes). Really it should trade off between all of these factors.
The challenge, of course, is that you don't know everyone's route ahead of time. Good prediction is important.
Essentially you're dynamically adjusting the bus schedule and route in response to spatial and temporal demand patterns. And with 10 passengers and no aisle[2], the maximum passenger density is similar to a bus (the "geometry" problem) So it seems like this can be more efficient than even a fixed bus schedule.
It seems like a decently large (100k+) multi-agent whole city transit simulation could settle the matter one way or the other. Who's plugging autonomous taxi routing into these?
This[3][4] looks promising. Accounts for induced demand. More publications on his website.[5]
[1] https://twitter.com/elonmusk/status/759153112993017856
[2] "Traffic congestion would improve due to increased passenger areal density by eliminating the center aisle and putting seats where there are currently entryways, and matching acceleration and braking to other vehicles, thus avoiding the inertial impedance to smooth traffic flow of traditional heavy buses." https://www.tesla.com/blog/master-plan-part-deux
[3] https://www.ethz.ch/content/dam/ethz/special-interest/baug/i...
[4] https://github.com/sebhoerl
[5] http://www.ivt.ethz.ch/en/people/person-detail.html?persid=2...