Jakob,
Thank you for getting back to me. Allow me to elaborate a bit more on what I am trying to achieve using SUMO.
Our research group is tasked with studying the impact of AVs and CAVs on freeway mobility and safety and we have chosen to employ SUMO for this purpose. One of the factors we study is the market penetration rate of these vehicle types. The simulation files that I sent in my previous email is related to a scenario where we have 20% CAV, 20% AV, and 60% Human-driven vehicles. We decided to use the ACC and CACC models of SUMO because they are the most updated and state of the art models developed so far for ACC and CACC and are actually based on experimental observations.
Could you please elaborate on your statement "I have not had enough experience in using these models to let me recommend them to your use"? Does this recommendation arise from the fact that these models a) are probably not able to faithfully represent the behavior of (C)ACC? or b) there is/are problem(s) with their implementation in SUMO?
Some comments/questions related to the proposed solutions are shown in red below. Please let me know what you think.
- replacing all carFollowModels with Krauss or W99 makes the issue disappear (flow on the rightmost lane roughly doubles) - ACC and CACC models were found to better represent the behavior of AV/CAV and thus we have decided not to model them using W99 or Krauss.
- using lcAssertive=1 instead of 2 also makes the issue disappear. In addition to simulating basic freeway segments, we are interested in merge and diverge segments. Having lcAssertive=1 for the merge and diverge scenarios makes things very bad for merging vehicles. These vehicles will not be able to enter the freeway and create an unstable condition at the merge influence area. That is why we chose to use lcAssertive=2 as it visually (observation of the GUI) represents the behavior close to what is expected in real life. It is worthy to mention that we also played with lcStatregic, LcCooperative, lcKeepRight, lcOvertakeRight, and lcLookaheadLeft. None of which significantly helped with the merging vehicle's problem.
- setting option --lanechange.overtake-right also makes the issue disappear. I will implement this in future simulations to see if it changes anything.
- replacing ACC with CACC looks a lot better than the original version but the drop is still quite large
Thank you,
Shoaib
Hello,
to be honest I'm not sure what exactly is going on in your scenario. However, here are some observations:
- replacing all carFollowModels with Krauss or W99 makes the issue disappear (flow on the rightmost lane roughly doubles)
- using lcAssertive=1 instead of 2 also makes the issue disappear
- setting option --lanechange.overtake-right also makes the issue disappear
- replacing ACC with CACC looks a lot better than the original version but the drop is still quite large
The ACC and CACC models are rather novel additions to the SUMO carFollowing model list (added within the TransAID project,
https://www.transaid.eu/) and are still undergoing changes. Bluntly speaking, I have not had enough experience in using these models to let me recommend them to your use.
There are also some known issues with these models:
I'm pretty sure the above flow issue is unrelated to the bug you cited as that was about merging effects and not specific to a particular car-following model.
regards,
Jakob
Hi Jakob and the team,
I am studying the impact of ACC and CACC equipped vehicles on basic segment capacity and simulated them in a mixed traffic stream where human-driven vehicles are modeled via W99. A (2 km) (3 lanes) basic freeway segment is simulated using the latest version of SUMO. A clear and significant drop in capacity (defined as the maximum 5-minute flow rate) is witnessed between the lanes. Moving from left most lane to the middle and from middle lane to the rightmost lane, the capacity decreases by as much as 50%. While a small (up to 15%) capacity drop is expected (happens in real life) between the left and right lanes, the drop in capacity obtained from SUMO is extremely large and not defensible.
The output of the simulation and other necessary files to run it are attached to this email for your reference.
- there is a known issue with the current dynamics model that can lead to a strong separation of vehicle speeds between the lanes (rightmost lane is very slow and leftmost lane is still at maximum speed). This lowers the observed capacity drop effect because there is no spill-over jamming effect to the left lanes. The speed seperation can be avoided with careful tuning of the lane-changing gap acceptance parameter (lcAssertive) distribution across the vehicle fleet but I anticipate that there will be model extensions in the future to adress this.
* Am I witnessing the issue because of a SUMO bug as highlighted by Jacok above?
* Are there any workarounds to solving this issue? I tried to carefully tune lcAssertive and other lane-changing parameters to no avail.
Thanks,
Shoaib
sumo-user mailing list
sumo-user@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://www.eclipse.org/mailman/listinfo/sumo-user
sumo-user mailing list
sumo-user@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://www.eclipse.org/mailman/listinfo/sumo-user
