On Collision Errors and Anomalies

  • Hello, everyone! Today I was playing and I thought I'd screw around a
    bit and have some fun playing with the game's technology. Therefore, I
    thought to myself, what if I parked myself in a way that would disrupt
    the other drivers? I was on the Road map Forza, which wasn't the widest
    of tracks, so this worked out well; eventually I had a car T-bone me, in
    a way similar to what you may see in the attached "bug3.png". However, a
    moment later, a second car came along! There I was, being slowly pushed
    by two cars down the track and giggling to myself like an idiot. Their
    not stopping, even after slowing quite a lot down, is a separate bug to
    report, but we have bigger fish to fry right now.
    Anyways, when I got bored of this silliness and decided to pull away,
    one of the cars did something... unique. As in "bug.png", it slowly
    raised up into the air before finally stopping, suspended 1-2 stories
    off the ground! I laughed to myself, and reloaded the map to see if it
    was just an isolated incident. As it turns out, it wasn't. I pulled some
    similar shenanigans next time, and stopped to observe and grab a
    screencap of the strange occurrence. Coincidentally, it was the same two
    cares that... M-boned me? Yeah, that works.
    The same car as earlier rose up into the air, exactly like it did the
    last time (albeit in a very different spot on the track) and started to
    slowly drift to the left.
    Instead of immediately attempting to duplicate it again, instead I
    watched it for a little while longer. What I didn't realize was that it
    was slowly drifting away, perpendicular to the way it was facing as it
    flew in the sky. It took a little while, but eventually as it hit the
    wall it disappeared without a trace! What was even stranger was the
    shadow left behind. You could see it at the right distance, but once you
    got too close it disappeared. You can see it, circled in bright red, on
    "bug2.png". I stared at it for a while, but nothing further happened
    that was interesting. I reloaded the map and tried again to duplicate it
    so I could get more data.
    Unfortunately, this time I didn't fare so well. While before it had been
    a yellow and lime-green car M-boning me, instead it was a blue and
    darker green car, the latter of which is actually pictured in
    "bug3.png". Similar results didn't occur, unfortunately.
    At the end, all I had left was these three screenshots and a headful of
    questions. Why did this just happen? Is it an act of God? AND HOW IS
    Thank you for your time, users and devs alike; I wish you all a merry
    Christmas and a happy new-year. If the attached images somehow don't get
    through, please notify me and I'll upload them and post links; you
    really can't appreciate the silliness of this bug without them.

    • Ok, this works all as intended;
      - The computer opponents run as skill level semi pro, so they receive damage on collisions
      - If the damage is too big (or in case of some other events), the car is considered as broken, and is taken out of the race by a virtual crane (lift up, move, lift down), if you run a level above rookie you can receive damage as well, see the damage bar in the counters.
      - The keep pushing behavior is to avoid locked up situations on the track, this forces a quick resolution, otherwise you could end up with a pile of cars sitting on the track just waiting the others to move first
      - To realism: To keep the project maintainable and small there are some shortcuts, we have no team of 50 people working full time on it and a budget of 20 Millions per release;-)

      So to keep TORCS maintainable, portable and performant we do not put a lot of complexity for obscure cases which should not happen, it is a design decision to handle such cases as simple as possible.

      Enjoy, Bernhard

  • Oh! Dreadfully sorry, I suppose I should have read the documentation a bit better. Still, it was very jarring! A good future goal that can be derived from this is to provide some sort of indication of what's happening.
    Just throwing around ideas here, but a good idea may be to add a simple particle effect, such as a small trail of smoke, to indicate damage.
    Alternatively, a simple model of the crane can come and pick the broken car up instead of the hand of god. I don't think either would be too difficult to implement, would it?
    While on some older computers the smoke could cause a minor performance hit, naturally there would be a way to omit that in the graphics settings.

  • Not your fault, this is nowhere explicitly documented:-)

  • Hello,

    Yes, there is a post explaining this, the car is taken out when the track is too dented, but did not find anything about a car "get inside the other."
    Also there is this post? Can you tell us where? Also, if this situation is expected, does not seem to be very realistic.

    What do you think?

    Thank you

    • It doesn't need to be realistic, because it is not an active part of the race. What I would like to add though is adding service cars (pacecar, trucks), so when an accident "kills" a car it stays where it is, the pace car goes out, and a service truck will collect/remove the wreck. Why do I care about this, but not about the crane: Service trucks/pacecar add to the race experience, it happens on the track. In opposite, the crane does not interact with the actual race, so it does not matter (so simplicity has priority).

  • Hello,


    A problem we encounter is the fact that when there is an accident and the car is on the contrary, it is possible that you lose the sense of the runway.
    We think it would be very difficult to warn the pilot that goes in the opposite direction, in Portuguese called "opposite", and this would bring undeniable advantages of gameplay.

    We put attached an example.

    Pedro Pereira, Miguel Marques

    • I don't think so, if you have a problem with the facing direction, you do actually not race. Beware, TORCS is not supposed to be a GT5/GTR/whatever clone, if you want these just go to the shop and buy it.

  • Hello,

    I feel in your words in a certain animosity, this is just an academic project and I'm just making suggestions. It was only to suggest improvements that could be accepted or not. If accepted very well, if they are not accepted in the same alright.

    Being an open source project, we do not need to purchase nor a clone of any game, is that we chose for our project and we want to improve it. That's it.

    We're sorry.

    Thank you.

  • No animosity, just a brief statement, if you would talk real life to me, you would know;-) I recommend to not interpret any feelings into written media, really.

  • Dear Bernhard,

    Another improvement that could be done, the track championship, the track itself, the tarmac is too similar to the side, as in the image, which causes some gameplay issues, since sometimes the car skids and not know why.
    You think you can do something about it, or rather was deliberate?

    thank you

    • It is intensional. The track is in the desert, and just the main track and pit lane (as you can see from the different texture) has a fine asphalt layer with good grip, the outer part is rough and much more bumpy and covered by some sand -> no grip. As you can see from the renewed parts in the turns, the grippy surface needs a lot of maintenance, this is because of high temperature gradients between day and night.

  • Dear Bernhard,

    After we analyzed several clues, found that when the car leans to the rails or boxes disappears, appearing only gauges with speed and rotations as in the image. Had you aware of this situation. We think it will not be difficult to solve.
    Here is a suggestion the total disappearance of the gauges, as the modified image and information "box leaning against the car," or else.

    Thank you.

  • Interesting, this might be a real bug now, good catch, I will review it next week, than we will see. I could even imagine that this is for some reason intensional, because just the gauges disappear and not all on screen data (so it doesn't seem to be a simple OpenGL state problem).

    Thank you for the report.

    • Ok, I looked into that, I misread your post, I assumed that the gauges are disappearing when contact happens, but this is actually your suggestion.

      I think this hint is not required, because you know where you are and from where you have come.

  • Dear Bernhard,

    Following our design discovery and subsequent publication on this forum, we found another small bug. On this track (see Annex) cars traverse (literally) when advertising for any reason, leave the track. We think that this error could be corrected through a barrier that would prevent the car from passing. I also thank your last post where valorizas our findings which gives us even more motivation to continue this project, further improving a game that, in our opinion, is already excellent in the spirit of open source.

    thank you

    • This is known, and it was a decision to handle this like that. When I implemented collision detection with pit walls and reworked it in general, I thought about that. The reason why I do not collision detect/response the distant markers is, that some "only XML" tracks are very simple, so there are no "landscape" features to recognize a braking point. So when cars would push these things away it would be harder to race the tracks.

  • Dear Bernhard,

    After we detected some errors "on the track", we also point out some aspects to the game off the track, so we think that there are tracks where the audience is "glued" ie, it is noted that no major concerns was placed Graphics .
    Moreover, there are no clues where the public does not make sense, in our opinion, because a race without spectators is like the circus without spectators. We think that, in a future revision of the game, these aspects can be improved.
    We put attached two examples of what we are saying.

    thank you

    • I see:-) You have to consider, that when you have time to care about this during your ride, you are actually not racing, otherwise you would be totally focused on the track and your opponents;-) Btw. I recommend you to watch some not so known racing series in television, then you will recognize that it is pretty common to race with empty stands (usually the case with GT* series, production car and entry level races, etc.).

      Better artwork is always welcome of course, so a more interesting track side makes sense, but I would not nail it down on spectators.

  • Dear Bernhard,

    As you know our group is trying to find some errors in TORCS in order to be corrected later. Again, we refer the fact that our posts are just academic suggestions involved in a project, nothing personal.

    However, we have noticed that you have validated our comments, you have not responded. The feedback to our posts is very important for us not only to show that we are engaged in the project, but also to guide us in the right direction. Your previous comments have been very assertive and encouraging.

    If you have some time, please review our posts.

    With our best regards.

    Thank you.

  • Sorry for the delay, but answering can take between 0 and 4 weeks, depends on my priorities.

  • Hello,

    It is with great pleasure that we realized we took the time to respond to our comments.
    Already now we have two question put to him. Since we started the project we have only raised questions in the forums. According realized, no one else has participated. We wanted to ask at what stage is the project: is the end, it will still take some more time ...?
    Moreover, according to Bernhard, that should this "stagnation"?

    thank you

  • I do not exactly understand the question? But this here is a bad place for communication for two reasons:
    - You use the "On Collision Errors and Anomalies" thread for your posts, so not many are interested to follow this -> wrong thread
    - This is a "default forum" and is nowhere officially announced, the "promoted" communication channel for this would be the torcs-users mailing list -> wrong place

    Best regards


  • Hello Bernhard,

    Some of our students have already given us feedback from usability/entertainment testing.
    Two of them identified the following errors:
    - When the car is in the sand, rear wheels do not follow the movement of the car, ie, the front wheels turn but the rear wheels remain motionless.

    • The car crosses the plates that are on the edges, this should not happen, it would be ideal, according to suggestions from students that the plates should fly through the air (This error has been identified previously by us).

    • When the car is in the sand is very difficult to get out of this. The car does not respond to commands when they press forward key to move, the car gives consecutive turns and starts reversing.

    We hope to have more information from the students and that they remain interested in this project.

    thank you

  • Dear Bernhard,

    We have supplied to our students/usability game tester the atached document in order for them to test TORCS according to Nielsen's heuristics adapted to games.

    Please have a look at the document and tell us if find necessary to make any adjustments regarding TORCS, you can make alterations or add more heuristics in order to better test TORCS.

    With our best regards,

    Pedro and Miguel