They’re definitely in sore need of muster points for the vehicles to hang out close to busy zones while not being called to a trip. Gotta imagine that’s what regional ops is focusing on.
Yea it's tough though since all of the demand is in high socio-economic areas (so RE is expensive) and high density (nowhere to legally park). The savvy Uber drivers will just park in red/waiting zones/driveways/private lots/etc but I know Waymo has gotten into a little trouble for this (no one wants a Waymo chilling out front all the time).
Wow, that's absolutely apalling. I really hope these technologies do not get wider adoption with those metrics because the effect on the roads feels pretty bleak.
Thanks, Harry! One thing worth adding is that higher fleet density isn’t free of challenges. When you pack more cars into the same area, the whole system suddenly needs much smarter geographic distribution. Otherwise you risk cars bunching up in popular zones while other areas run thin, which actually creates new deadheading as the fleet constantly rebalances itself. In other words, density helps, but only if the network learns how to spread those extra vehicles intelligently.
True. Uber does a good job of this by using surge pricing and heat maps for drivers so they basically use algorithms to coax drivers into certain areas and balance out density.
I see, thanks. I was hoping to make a more detailed analysis of peak-hour vs. off-peak deadheading, but it looks like you've already made the most detailed breakdown of VMT possible with this very limited data.
They’re definitely in sore need of muster points for the vehicles to hang out close to busy zones while not being called to a trip. Gotta imagine that’s what regional ops is focusing on.
Yea it's tough though since all of the demand is in high socio-economic areas (so RE is expensive) and high density (nowhere to legally park). The savvy Uber drivers will just park in red/waiting zones/driveways/private lots/etc but I know Waymo has gotten into a little trouble for this (no one wants a Waymo chilling out front all the time).
Wow, that's absolutely apalling. I really hope these technologies do not get wider adoption with those metrics because the effect on the roads feels pretty bleak.
Thanks, Harry! One thing worth adding is that higher fleet density isn’t free of challenges. When you pack more cars into the same area, the whole system suddenly needs much smarter geographic distribution. Otherwise you risk cars bunching up in popular zones while other areas run thin, which actually creates new deadheading as the fleet constantly rebalances itself. In other words, density helps, but only if the network learns how to spread those extra vehicles intelligently.
True. Uber does a good job of this by using surge pricing and heat maps for drivers so they basically use algorithms to coax drivers into certain areas and balance out density.
Yup. Will be interesting to watch how this plays out. Thanks again for sharing.
Great analysis.
When I look at the trips files from CPUC, all of the VMT columns are redacted. Where are you getting the P1/P2 VMT?
It's a bit confusing but go here and download Waymo's 'Deployment Program Quarterly Reports' -
https://www.cpuc.ca.gov/regulatory-services/licensing/transportation-licensing-and-analysis-branch/autonomous-vehicle-programs/quarterly-reporting
Here's my spreadsheet with the data - https://docs.google.com/spreadsheets/d/109epdBMshkEnzhtDR9U_5gN18CAdpbzg/edit?gid=498277451#gid=498277451
Video explainer - https://www.loom.com/share/a3ef428977304e068dd973b909d6ff0b
I see, thanks. I was hoping to make a more detailed analysis of peak-hour vs. off-peak deadheading, but it looks like you've already made the most detailed breakdown of VMT possible with this very limited data.