search

arpg-sophisticated / ar-tu-do

ROS & Gazebo project for 1/10th scale self-driving race cars
18 stars 5 forks source link

Stable in Gazebo #131

Open ghost opened 4 years ago

ghost commented 4 years ago

WF2 and 5 fails in simulation, please ensure, that both, or at least wf5 can run this track.

jhbruhn commented 4 years ago

Fails to start or due to collisions?

ghost commented 4 years ago

crash bandicot style ....

even on low speed, the second 90° is deadly ... just run gazebo from current wallfollowing-improvements branch, you'll see

Am Mi., 12. Aug. 2020 um 13:19 Uhr schrieb Jan-Henrik Bruhn < notifications@github.com>:

Fails to start or due to collisions?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/arpg-sophisticated/ar-tu-do/issues/131#issuecomment-672811491, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHFWFWN5Z6534NFDLUCC2ETSAJ3CLANCNFSM4P4IFWPA .

jhbruhn commented 4 years ago

First improvements can be seen by setting the weight of the car in the simulation to 9kg (instead of 3.5kg). What value is the real value here? I have 9kg in my head but I don't know if that's the actual weight of the car. Otherwise, at least for WF5, this is a thing of properly tuning the parameters for the track. I have done some testing and got nearer to a good driving performance by changing parameters, but this only makes sense as soon as the algorithm is "finished" -> parameter impacts don't change anymore. About WF2: I am unsure about how well parameter tuning works here, still have to try.

Rothu commented 4 years ago

WF2 doesn't run properly on wallfollowing-improvements anymore, because I added a speed correction factor in car_controller to get the intern speed value closer to reality.

jhbruhn commented 4 years ago

Can confirm, adding that factor to wallfollowing2 indeed fixes the simulation of it.