But when I upload the same solution, it gives an error stating Solution is not feasible! Feasibility by planning problem id: {30274: 'Not Feasible!'}. What could be the possible issues?
Hi Shivesh, the vehicle might also collide with obstacles or leave the road. All of these checks are performed in commonroad_dc.feasibility.solution_checker.valid_solution().
Hi Moritz,
Thank you for the reply. I assumed that Solution is not feasible only comes up when the solution trajectory does not comply with the car kinematic/dynamics model. Because collision with obstacles show There is a collision between the scenario obstacles and the ego vehicle in planning problem solution and laneboundary violation shows There is a collision between lanelet boundaries and the ego vehicle in planning problem solution.
You’re actually right, there are already different error messages and this error actually means the kinematic feasibility. Could you send the solution file to moritz.klischat@tum.de?
It might be that you have another version of scipy than the one which is installed on the server. What is your scipy version? Check with pip show scipy.
Hello, Shivesh sent me his solution yesterday and I didn’t find the problem yet.
If it’s related to a problem on the server, we will re-run the feasibility check on the failed solutions.
I will keep you updated here.
At the end of the lane change at about 0:06, the motion seems a bit abrupt which could be the reason for the failure.
However, that doesn’t explain why there are different results on the server. It could be related to numerical problems with the solver that is involved in the feasibility check on different machines. We adjusted the solver settings on the server and re-evaluated the solutions. Perhaps you can check if your affected solutions are rated as feasible now?
I checked your submissions: In this case the error message was not related the kinematic feasibility, but indicates a collision with other vehicles. That might be related to an older version of the sumo interface you are using to create the solution. This can result in slightly different trajectories of other vehicles. So when replaying your solution on the server, it could result in a collision. Are you using the latest version? https://pypi.org/project/sumocr/
Thank you a lot!
I updated my sumocr and I guess the local collision-checker is currently congruent with that on the website. But mismatches still exist in the results of feasibility check . My latest two submissions are locally checked to be feasible but failed on the website.