Hi All,
My project is a carpooling application (web–> Tomcat + PostgreSQL and mobile --> Android) like Blablacar but for a short distance.
Our Search section works but we would to improve the routing matching algorithm.
Users (passengers) have to fill this required fields when they search a trip:
Origin,
Destination (postigis coordinates),
max waiting time
max walking distance
time window (day and hour).
The users (Drivers), instead, offers their rides filling this fields:
Origin,
Destination,
seats available,
trip preferences (smoke, music, etc… number from 1 to 5),
time window (day and hour).
Every user’s profile contains trip preferences by default and can be changed every moment.
Our goal is to return an optimal list of trip offers with trajectory (path), time window and especially preferences when a user search a particular trip.
Hi,
Interesting use case! If I were you, I would start and test with a simpler problem. For example, when it comes to passengers, you can accomodate max waiting time and time window (you would just model both with a time window), and when it comes to drivers, you can easily accomodate capacity (seats) and time windows. However, depending on the number of drivers, you would need a fast pre-processing to identify promising drivers. If this works for you, I assume that you can consider preferences (smoke, music) in soft-constraints as well.
Best,
Stefan
You should probably have to edit the Objective function in order to only permit a little deviation from the driver route.
To know if a passenger is on the way, you may have to use a preprocess able to give you an intersection between the walk area and the vehicle route . In order to have new coordinates or time/distance matrix conciliable with the new Objective function.
Thanks for reply, guys.
I’ll try to explain better how our application works and what we would like to do with jsprit.
In my last post I explained how the search section (web and mobile application) works.
I forgot to explain what it do when a driver offers a ride.
Basically, when a driver offers a ride all the informations (origin, destination, trip preferences, time window and others not mandatory) are saved into a DB with its trajectory into a particular table.
The info into this particular trajectory (path) contains also info about user (the driver).
The current algorithm give us an intersection between the walk area and the vehicle route but sometimes this intersection are unlikely. Our processing occurs into DB through a complex query with many constraints. How jsprit can help us to improve our algorithm?
Where to start?