Package org.matsim.contrib.parking.parkingsearch

Thoughts related to parking:
  • The agents to not learn routes any more from day to day: routes are completely computed on the fly.
  • How does this pick up congestion? At least in theory, the router may have a Travel Time that comes from the past. In the sense of a Nash equilibrium, this seems imperative.
  • It is not possible to learn routes from day to day ... since the car may be parked at different starting points every day. So one thing that could be learned is some kind of private mental map. Instead the approach now uses a global mental map ... an ok approximation.
  • There was the "Milano" problem: The router may be far off when planning several hours into the future. Could be addressed by regular (e.g. every 15min) replanning.
  • The agent walk-teleportes from the activity facility to the parking facility. Should she eventually consider network walk, we need to be clear about access-walk -- network-walk -- access-walk??? -- car.
Author:
jfbischoff, nagel