Stuck and abort agents in car sharing API

[s-atac]

Hello,

I am using the carsharing API in MATSim. For testing purposes, I consider the Sioux Falls network with 84110 agents. For the carsharing system, I arbitrarily place 24 stations at the intersection of main roads, with each station having 500 cars and 250 additional free parking available at the beginning of the day. I assume that each agent has access to the carsharing system. Apart from carsharing, walk, bike, pt, and car are also available modes to the agents.

I would like to use the events output file for my research; however, I have been having problems with stuck and abort agents. I explain some analysis that I have done below:

I observe that there are three types of incidences: 1- The agent performs some activities, but then stuck and abort, (13%) 2- The agent performs all the planned activities, (34%) and 3- The agent is directly stuck and abort before performing any planned activity (52%). Here, the agents that are stuck and abort at some point of their activities sums up to ~65%, whereas the agents who can complete their daily routine is only at ~34%. This does not sound very logical to me and I do not know how to interpret this result if I want to use the events output file for further research.

To see whether these values would change, I tried to play with flowCapacityFactor, storageCapacityFactor, stuckTime, and the weights of different strategies (ReRoute and BestScore). However, the results change very slightly.

As a last note, according to the Stuck Agent Analysis in Via, I see that most of the agents (85% of the stuck and abort agents, 54% of all population) get stuck at the beginning of the simulation.

I would really appreciate your help on interpreting these situations, and any solutions if possible.

Thank you.

[balacmi]

Hi Selin,

Could you provide us with the following information please:

• logfile of the simulation where this occurs
• does this happen without carsharing?
• does this happen already in iteration 0 or later on?
• how does your input population look? Is it a default one for SiouxFalls or do you make any adjustments?
• please mind that agents that want to use carsharing but cannot find a station or vehicle within a certain distance from their departure point are by default aborted and heavily penalized

[s-atac]

Hi Miloš,

Thank you for your reply.

• The log file is attached. run1.logfileWarningsErrors.log
• Yes, it does. According to the Stuck Agents Analysis in VIA, iteration 0 results in 51337 stuck agents (44333 of them at hour zero, i.e., ~86%) where iteration 10 has 53374 stuck agents (45397 of them at hour zero, i.e., ~85%).
• I do use the default Sioux Falls population only with an adjustment of adding membership to carsharing for each agent.
• I understand that, however, the stuck and abort situation does not really change even if I increase the available number of cars and parking spots.

Thank you.

[balacmi]

Hi Selin,

Can you please provide the full log and not errors and warnings logfile.

If this is happening in iteration 0 it should have nothing to do with carsharing. It must be some other configuration that you have.

I am not familiar with stuck agents analysis in VIA. Can you tell me are these passenger agents or can these be pt agents?

Could you verify in the events file which agents are stuck and when?

[s-atac]

Hi,

Sorry, here is the full log: run1.logfile.log

I see. Do you have any idea on what it could be? I have tried with more number of stations and free parking spots, however it doesn't change. As I said before, I also played with some parameters that I thought might be relevant.

I am not sure to understand what passenger agents and pt agents refer to, but when I analyse the events file, I see that some agents are stuck right away (even before any event log), some agents are stuck right after Rental Start, and some are stuck at their daily activities, e.g. secondary.

[balacmi]

Hi Selin,

I see from your log file that you did change your input plans by assigning carsharing mode to a lot of trips. Can you please tell us what you did there?

I would suggest that you start with nobody using carsharing and then let the simulation go towards equilibrium. Probably at the start you have a large disbalance between supply and demand. Also your strategies have strange weights. I would suggest that SelectExpBeta is used in 80% of the cases and the rest together in 20%. I would also not use BestScore. Currently, you have too much innovation and that is not allowing the simulation to go towards equilibrium.

[s-atac]

Hi Miloš,

Thank you for your explanatory reply. I tried to involve only walking and carsharing in the system, and I think that's why I am having this problem and now it makes sense to me that it cannot reach to equilibrium. Regarding the strategies, I have tried so many things, some of them did not make much sense but I just wanted to see whether there is a reasonable change.

Thank you for your help, I'll see if I can resolve this issue by decreasing the number of carsharing trips.