From: Daniel K. <d.k...@go...> - 2010-10-20 20:43:16
|
Dave, I just submitted something that should at least reduce the blockings in front of intersections. Would you mind to test it? The problem of vehicles entering the intersections though no place behind is available is not yet solved. sincerely, Daniel 2010/10/1 Daniel Krajzewicz <d.k...@go...>: > Hello Dave, > > You may bet we're keeping you in mind the last days. Yes, the > behaviour you describe is known to us and actually, we suffer the same > problems. Patching problems in lane-changing is what we are trying to > do right now. > > Please find below some comments and hints. > > 2010/9/28 Dave McKenney <dav...@ho...>: >> I never heard back about this. Does anybody have any ideas? I think I may >> have done something wrong with specifying prohibitions. >> >> I also removed the ability to turn right on red and that seemed to fix a bit >> of the problem, but not all of it. >> >> ________________________________ >> From: dav...@ho... >> To: d.k...@go... >> CC: sum...@li... >> Subject: RE: [sumo-user] Blocked Intersections >> Date: Sun, 12 Sep 2010 22:00:47 -0400 >> >> Okay, I updated and used the SVN version and it has alleviated the problems >> a little bit. I am still getting some weird behaviour however. I'm not sure >> if I can attach pictures, but I will try either way. Below is a description >> of what is happening in the images. >> >> stopped1.JPG: The vehicles have backed up into the intersection so vehicles >> cannot pass through it in the opposite direction. In the last message you >> said you hadn't seen it before, so I included this even though it isn't the >> weird happenings really. >> > > The vehicles are trying to keep the intersection open, but the > algorithm obviously fails. > By now, vehicles count the free place in front of them together with > the number of vehicles which will approach the lane. The problem is > that it is not possible to look at the current time step only, as > vehicles may for instance stop in front of an intermediate > intersection. Also, if vehicles are driving, it may be that they pass > an intersection giving further place. Not counting this yields in > vehicles halting until they have additional space. All in all, that's > pretty complex. Indeed, this is not a lane-change problem, but yes, we > have to address this in the next time - before the next release - too. > BTW, this misbehaviour is probably more frequent in your scenario > because of the very small vehicle size you use. > > >> stopped2.JPG: The vehicles in the northbound direction have a green light >> (the plan for these connections is all 'G'). However, they just sit there >> and don't move. The right turning lane in the westbound road has a 'g' light >> to allow for right-turns-on-red if no other vehicles are coming. I'm >> thinking this may be what is causing the issue, but it doesn't make sense >> that it would also stop vehicles that wish to turn left (the 2 left-most >> lanes in the northbound road are left turning lanes). >> >> stopped3.JPG: Like above, only both north/south is green. I understand the >> vehicles don't turn left because of approaching vehicles, but the straight >> vehicles also aren't moving. >> > > The reason for standing is that vehicles are blocking each other. If > you right-click on a vehicle, you may enable the "show current route" > visualisation. You will probably notice that a vehicle stays on right > lane, but wants to go straight what's not possible from the lane it is > on and the one on the lane left to him wants to turn right. So they > block each other. In some earlier versions, such vehicles were simply > swapped. Surely not very realistic, but it works. To the current SVN > version, I added the option --lanechange.allow-swap. I suppose this > should help and as your vehicles have all the same length, this > probably should help. > The real solution needs some steps: > a) solving blocking behaviour as soon it occures in a clean, realistic > way (one vehicle lets the other use his link to the next lane) > b) try to avoid such situations at all by improving the lane-changing > interaction > c) try to avoid such situations by better lane-2-lane connections > > As told, we are currently working on this, but you see that the this > is quite complex. We are planning to release a new version, soon, and > will keep you informed if something about lane-changing and keeping > intersections free changes. > > sincerely, > Daniel > >> empty.JPG: This just has the names of the edges on it so the things below >> make more sense. >> >> Also, this oddly does not occur at all times, it just starts later on in my >> simulation. I had to specify connections myself because they were not >> implied the way I wanted them. After specifying them however, different >> lanes were interfering (prohibiting) others that you wouldn't expect, so I >> defined the following prohibitions myself. >> >> <prohibition prohibitor="Elgin-Slater-Albert_0->Elgin-Albert-Queen_0" >> prohibited="Albert-East-Elgin_0->Elgin-Albert-Queen_0" /> >> <prohibition prohibitor="Elgin-Queen-Albert->Albert-Elgin-Metcalfe" >> prohibited="Elgin-Slater-Albert->Albert-Elgin-Metcalfe" /> >> <prohibition prohibitor="Elgin-Queen-Albert->Elgin-Albert-Slater" >> prohibited="Elgin-Slater-Albert->Albert-Elgin-Metcalfe" /> >> <prohibition prohibitor="Albert-East-Elgin_0->Albert-Elgin-Metcalfe_0" >> prohibited="Elgin-Queen-Albert_0->Albert-Elgin-Metcalfe_0" /> >> >> >> >> Maybe I did something wrong, I'm not sure. I can now at least run an entire >> simulation without complete gridlock, but with things not working correctly >> any results I create won't mean anything. >> >> Anyway, if you can give me any advice/help it would be great. If not, at >> least you are aware there are some issues (whatever they may be...). >> >> Thanks again, >> >> -Dave >> >> >> >> ________________________________ >> Date: Sun, 5 Sep 2010 22:29:38 +0200 >> Subject: Re: [sumo-user] Blocked Intersections >> From: d.k...@go... >> To: dav...@ho... >> CC: sum...@li... >> >> Hello Dave, >> >> yes, the problem is pretty known to us, too. Are you using 0.12 or the SVN >> version? I think the SVN version's behaviour is less severe... Actually, >> that's the only help I have so far. >> >> There are several issues >> a) in the imported networks we often find nearby intersections which jam >> when vehicles from opposite directions want to turn left - they jam each >> other >> b) the intersections are not kept free for opposite traffic properly >> >> I am a littel bit surprised that opposite traffic can not pass. I have not >> seen this before. Tangling the point that a vehicle is already prohibiting >> foes if standing only a bit on the intersection seems to require either >> massive geometry computing maths during the simulation or having large >> look-up table which would blow the network size. >> >> sincerely, >> Daniel >> >> >> >> sincerely, >> Daniel >> >> 2010/8/28 Dave McKenney <dav...@ho...> >> >> Is there any (preferably easy) way to stop vehicles form blocking >> intersections? >> >> I have a simulation that has high volumes at times, and I have been having >> problems with vehicles trying to move across the intersection only to be >> blocked because the edge they are moving to is backed up. I would like to >> stop them from moving across in this case, so vehicles don't become >> deadlocked. Even if vehicles traveling across the intersection in the >> opposite direction could bypass these, it would be helpful. >> >> Thanks, >> >> -Dave >> >> ------------------------------------------------------------------------------ >> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program >> Be part of this innovative community and reach millions of netbook users >> worldwide. Take advantage of special opportunities to increase revenue and >> speed time-to-market. Join now, and jumpstart your future. >> http://p.sf.net/sfu/intel-atom-d2d >> _______________________________________________ >> sumo-user mailing list >> sum...@li... >> https://lists.sourceforge.net/lists/listinfo/sumo-user >> >> > |