population size effect on DRT pooling

[elyh1992]

Hi,

I have two questions about DRT especially in the case of pooled DRT. In my simulation, I have a DRT service with the possibility of pooling up to 4 passengers in one car but when I look into the DRT occupancy time profile most of the rides are having only one passenger and the percentage of rides with 2, 3, or 4 passengers are very rare. As in MATSim we usually use a percentage of the population in simulation (in my case 1 percent of the population), I wanted to know how effective the scaling down of the population is on the possibility of successful pooling? Also, I wanted to know is there a way to increase the possibility of pooling in the DRT extension (any specific configuration that pushes more people towards pooling their ride)?

Please let me know if you have any suggestions about increasing the possibility of a successful match or you know of any work or paper related to this matter. Thank you in advance for your answer.

[tschlenther]

Hello,

the alpha and beta parameters (for dispatch) have major influence on the pooling likeliness. You can tune the acceptable detour by adjusting them.

In any way, a one percent scenario is probably too small to properly model the drt demand and performance. Which scenario are you investigating?

Please have a look at this paper (VSP Working Paper 20-36), where the scaling effects of drt were investigated. It suggests that very small sample sizes are not really suited to model drt use cases....

[elyh1992]

Hello,

the alpha and beta parameters (for dispatch) have major influence on the pooling likeliness. You can tune the acceptable detour by adjusting them.

In any way, a one percent scenario is probably too small to properly model the drt demand and performance. Which scenario are you investigating?

Please have a look at this paper (VSP Working Paper 20-36), where the scaling effects of drt were investigated. It suggests that very small sample sizes are not really suited to model drt use cases....

Hi,

Thank you for your helpful response. I am using the configuration that does not reject a request when the maximum wait time and travel time are violated. Are alpha and beta parameters effective on pooling likeliness even in this case?

I am investigating scenarios that I have two different DRT services one with the possibility of pooling and one without the possibility of pooling and I am trying to design policies that increase the uptake of pooled ride services. at first I had all modes including pt,car,ride,bike and walk and I added two drt modes and let people to switch to these modes if they wanted but I noticed that one percent population is not enough for properly modeling the drt demand. So I decided to replace all car and ride modes with these two DRT modes. However still a great percentage of drt rides are only with one passenger. I guess this might be because of the fact that anyways in 1 percent sample there aren't many people with matching origin and destinations so even if the whole 1 percent sample use drt as their mode of transport, still there aren't enough sharable rides. I think I need to increase population size at least to 5 or 10 percent. What do you think about it? the problem is that with large sample sizes the run time is very long. Do you suggest using DRT speed-up extension? I checked the paper that you suggested and I found it really helpful. Thank you for your help.

[tschlenther]

Thank you for your helpful response. I am using the configuration that does not reject a request when the maximum wait time and travel time are violated. Are alpha and beta parameters effective on pooling likeliness even in this case?

yes, they should be.

I think I need to increase population size at least to 5 or 10 percent. What do you think about it? the problem is that with large sample sizes the run time is very long. Do you suggest using DRT speed-up extension?

yes, you can consider using the speed-up. As mentioned above as well as in the paper, the results depend on demand density. This is why, like you stated yourself, setting all legs in a 1 percent sample has not the same effect as simulating 10 or 20 percent. My suggestion is to a demand sampleof 10 percent or more and to use the drt speed up extension, if necessary.

[Areza622]

@elyh1992 Hi, I want to do the same simulation as you did, I know how to run drt using "matsim maas" but I am interested to do like what you have done(letting people choose between different modes), would you please help me how should I do that and which project of matsim should I use?