Hello, during the holiday seasons, responses may take longer. In the scenario under discussion, the junctions beyond the intersection must be modeled with zipper type to achieve the correct lane-choice ahead of the intersection (as you yourself noticed). I suspect that there is a bug in the interaction of the action-step model with zipper junctions that leads to the observed collisions. It may take some time until this is fixed. (tracked at https://github.com/eclipse/sumo/issues/4330)
One possible work-around for the scenario would be to shorten the routes so they end just before the zipper junction. This will preserve all relevant traffic at the intersection, prevent adverse effects on lane choice as well as circumvent the collisions. regards, Jakob 2018-07-12 13:40 GMT+02:00 sung hyun Jang <xdd...@gmail.com>: > Dear Jakob Erdmann > > Thank you for your reply. I followed your three recommendations. I > attached file to this Email. > > Actually, I am using* the SUMO file from DLR (Tostmann platz)* to compare > with a network that I created with bike and pedestrians. I sent Email to > you directly, but you didn't answer.. > so I am asking you by mailing list, again. > > *I changed detectors and decoupled decision intervals from step length, > because real controller should check at 100 ms* *interval. * > > Collision and teleporting occurs in narrow connection point from two lanes > to one lane in *NS and SN direction.* > > The result of change based on your recommendations is the following. > > First, for *run.sumocfg, *the change junction type to "Zipper" was > already applied at the original net file from DLR. Only change is > step-length and action-step-length. > > Second, for *run_EdgeConnection.sumocfg,* *one of the lanes does not > have a connection to the next edge.* Only this case, there is no > collision, but the problem is that vehicles try to change lanes at the > beginning of edges even if still there is another edge left before narrow > point from two lanes to one lane. For real condition, it doesn't make sense. > > Third, for *run_Priority.sumocfg,* *two connections from the same edge to > the same target lane where one connection has priority over the other one.* > I set a priority on one of connections with Target(pass) option. The other > connection has target (light Green color). It doesn't look different from > run.sumocfg (Zipper merging). > > I wonder if we apply action-step-length for a network, collisions would > often occur.. and if we should do something to prevent the situation. > > I really hope you can give a solution! I am looking forward to your reply. > Thank you for your time in advance. > > Best regards, > Sunghyun Jang > > > On 8 July 2018 at 23:10, Jakob Erdmann <namdre.s...@gmail.com> wrote: > >> There are three "appropriate" ways to model this type of situation and >> also "inappropriate" ones. >> Which of the following did you use? >> 1) one of the lanes does not have a connection to the next edge (vehicles >> are forced to change lanes to continue) >> 2) junction type 'zipper'. vehicles perform zipper-merging >> 3) two connections from the same edge to the same target lane where one >> connection has priority over the other one >> >> Actually, all of these should be collision free but the action-step >> extension is still quite new so there could be a bug. >> An inappropriate way to model the situation would be with two connections >> that are in conflict but without a priority order. >> If you are using one of the first three, please send me the scenario so I >> can determine the source of the problem. >> >> regards, >> Jakob >> >> 2018-07-07 16:28 GMT+02:00 sung hyun Jang <xdd...@gmail.com>: >> >>> Dear SUMO users >>> >>> I have a question about collision and teleporting vehicles in simulation. >>> >>> Actually, I decoupled the decision interval from the simulation step >>> length. >>> >>> step-length 0.1s / action-step-length 1s >>> >>> After that, I have a lot of collision and teleporting vehicles at the >>> connection points from a link with two lanes to a link with one lane. >>> Before, I had no collision and teleporting vehicles. >>> >>> In order to solve the situation, I tried several things. >>> >>> First, I changed the shape of junction, for example, little large and >>> long in net file. >>> Second, in config file, I added command line. >>> <collision.mingap-factor value="0"/> >>> Third, for <vType>, I changed tau (reaction time) from default to 2s.. >>> and I reduced traffic flow 10~20%.. >>> >>> But still I have 5 ~10 teleporting vehicles because of collision. >>> >>> I wonder if there is any way to avoid the situation.. >>> >>> I am looking forward to your help. Thank you in advance. >>> >>> Best regards, >>> Sunghyun Jang >>> >>> >>> >>> _______________________________________________ >>> sumo-user mailing list >>> sumo-user@eclipse.org >>> To change your delivery options, retrieve your password, or unsubscribe >>> from this list, visit >>> https://dev.eclipse.org/mailman/listinfo/sumo-user >>> >>> >> >> _______________________________________________ >> sumo-user mailing list >> sumo-user@eclipse.org >> To change your delivery options, retrieve your password, or unsubscribe >> from this list, visit >> https://dev.eclipse.org/mailman/listinfo/sumo-user >> >> > > _______________________________________________ > sumo-user mailing list > sumo-user@eclipse.org > To change your delivery options, retrieve your password, or unsubscribe > from this list, visit > https://dev.eclipse.org/mailman/listinfo/sumo-user > >
_______________________________________________ sumo-user mailing list sumo-user@eclipse.org To change your delivery options, retrieve your password, or unsubscribe from this list, visit https://dev.eclipse.org/mailman/listinfo/sumo-user
