For many scenarios there doesn’t seem to be a collision free path.
If there’s no collision free path, then most often there isn’t a successor from the start node that passes the self.is_collision_free(path_translated) test.
Am I missing something or are many scenarios unsolvable?
Hi,
I believe my colleague generated the scenarios by removing a vehicle from simulation results. So in principle they should all be solvable. Can you maybe specify some of the scenarios which you suspect are not solvable?
For a quick check, you can see the scenarios solved by other students on the leaderboard here: https://commonroad.in.tum.de/challenges/7152aa1b-cdf6-4155-b036-793935bef331 (click on arrows on the left). If a scenario is on the list, it has then passed through checks on the server.
There is a student with 370 solutions so far. If you don’t see any of the suspected scenarios on the list, you can send me some of those (edmond.irani@tum.de) so I can have a closer look at them.
Hi Edmond,
I followed your advice and checked whether other people were able to solve the scenarios for which I get an empty frontier. Most of the scenarios don’t have a solution, but some, such as USA_Lanker-1_10_T-1(https://commonroad.in.tum.de/scenarios/USA_Lanker-1_10_T-1/benchmarks), do. You are one of the people that solved it. Is it possible that I need to change the motion primitives to solve those scenarios?
List of scenarios for which I get an empty frontier right in the first step:
AM_Ramp-1_1-T-11-1_T-1T-1
BEL_Putte-10_1_T-1
DEU_BadEssen-1_5_T-1
DEU_Flensburg-64_1_T-1
DEU_Guetersloh-55_2_T-1
DEU_Ibbenbueren-8_6_T-1
DEU_Lohmar-19_1_T-1
ESP_Ceuta-11_1_T-1
ESP_Inca-3_2_T-1
ESP_Toledo-8_1_T-1
FRA_Anglet-2_1_T-1
HRV_Pula-2_2_T-1
ITA_CarpiCentro-1_1_T-1
ITA_Foggia-12_1_T-1
ITA_Foggia-19_1_T-1
USA_Lanker-1_10_T-1
USA_US101-5_1_T-1
USA_US101-5_2_T-1
ZAM_Ramp-1_1-T-1
Full list of scenarios for which I eventually get an empty frontier:
ARG_Carcarana-12_2_T-1
ARG_Carcarana-8_4_T-1
BEL_Aarschot-6_1_T-1
BEL_Putte-10_1_T-1
BEL_Putte-14_1_T-1
BEL_Putte-14_2_T-1
BEL_Putte-14_3_T-1
BEL_Putte-16_2_T-1
BEL_Putte-16_6_T-1
BEL_Putte-17_1_T-1
BEL_Putte-2_3_T-1
BEL_Putte-4_1_T-1
BEL_Zaventem-2_5_T-1
DEU_A99-1_1_T-1
DEU_A99-1_2_T-1
DEU_B471-1_1_T-1
DEU_Backnang-7_3_T-1
DEU_Backnang-7_4_T-1
DEU_BadEssen-1_5_T-1
DEU_BadWaldsee-4_1_T-1
DEU_Flensburg-15_1_T-1
DEU_Flensburg-17_1_T-1
DEU_Flensburg-19_1_T-1
DEU_Flensburg-24_1_T-1
DEU_Flensburg-38_1_T-1
DEU_Flensburg-40_1_T-1
DEU_Flensburg-50_1_T-1
DEU_Flensburg-58_1_T-1
DEU_Flensburg-64_1_T-1
DEU_Flensburg-74_1_T-1
DEU_Flensburg-75_1_T-1
DEU_Flensburg-83_1_T-1
DEU_Flensburg-84_1_T-1
DEU_Flensburg-92_1_T-1
DEU_Guetersloh-10_4_T-1
DEU_Guetersloh-1_3_T-1
DEU_Guetersloh-26_3_T-1
DEU_Guetersloh-28_1_T-1
DEU_Guetersloh-28_2_T-1
DEU_Guetersloh-32_2_T-1
DEU_Guetersloh-5_1_T-1
DEU_Guetersloh-55_2_T-1
DEU_Guetersloh-9_3_T-1
DEU_Hennigsdorf-2_6_T-1
DEU_Hennigsdorf-9_4_T-1
DEU_Hhr-1_1
DEU_Ibbenbueren-10_3_T-1
DEU_Ibbenbueren-5_2_T-1
DEU_Ibbenbueren-8_6_T-1
DEU_Lohmar-10_1_T-1
DEU_Lohmar-12_1_T-1
DEU_Lohmar-13_1_T-1
DEU_Lohmar-18_1_T-1
DEU_Lohmar-19_1_T-1
DEU_Lohmar-2_1_T-1
DEU_Lohmar-25_1_T-1
DEU_Lohmar-26_1_T-1
DEU_Lohmar-31_1_T-1
DEU_Lohmar-38_1_T-1
DEU_Lohmar-42_1_T-1
DEU_Lohmar-55_1_T-1
DEU_Lohmar-58_1_T-1
DEU_Lohmar-59_1_T-1
DEU_Lohmar-61_1_T-1
DEU_Lohmar-65_1_T-1
DEU_Lohmar-66_1_T-1
DEU_Lohmar-70_1_T-1
DEU_Luckenwalde-1_1_T-1
DEU_Muc-2_1_T-1
DEU_Muc-2_2_T-1
DEU_Muehlhausen-14_1_T-1
DEU_Reutlingen-1_1_T-1
DEU_Reutlingen-3_1_T-1
DEU_Rheinbach-1_2_T-1
DEU_Speyer-2_1_T-1
DEU_Speyer-3_4_T-1
ESP_Ceuta-11_1_T-1
ESP_Ceuta-5_1_T-1
ESP_Inca-3_2_T-1
ESP_SantBoideLlobregat-22_3_T-1
ESP_SantBoideLlobregat-4_1_T-1
ESP_SantBoideLlobregat-6_1_T-1
ESP_Toledo-8_1_T-1
ESP_Toledo-8_3_T-1
FRA_Anglet-2_1_T-1
FRA_Anglet-4_5_T-1
FRA_Anglet-4_6_T-1
FRA_Anglet-5_2_T-1
GRC_Perama-1_1_T-1
GRC_Perama-6_1_T-1
HRV_Pula-18_3_T-1
HRV_Pula-2_2_T-1
HRV_Pula-7_4_T-1
ITA_CarpiCentro-1_1_T-1
ITA_CarpiCentro-5_3_T-1
ITA_Empoli-10_1_T-1
ITA_Foggia-12_1_T-1
ITA_Foggia-19_1_T-1
ITA_Segrate-1_1_T-1
ITA_Siderno-2_2_T-1
ITA_Siderno-3_4_T-1
USA_Lanker-1_10_T-1
USA_Lanker-1_12_T-1
USA_Lanker-1_4_T-1
USA_Lanker-1_5_T-1
USA_Lanker-2_20_T-1
USA_Lanker-2_2_T-1
USA_Lanker-2_5_T-1
USA_Lanker-2_6_T-1
USA_Lanker-2_7_T-1
USA_Lanker-2_8_T-1
USA_Peach-3_1_T-1
USA_Peach-3_2_T-1
USA_Peach-3_4_T-1
USA_Peach-4_3_T-1
USA_US101-10_1_T-1
USA_US101-10_4_T-1
USA_US101-11_1_T-1
USA_US101-11_4_T-1
USA_US101-1_1_S-1
USA_US101-1_1_T-1
USA_US101-12_1_T-1
USA_US101-12_4_T-1
USA_US101-13_1_T-1
USA_US101-14_1_T-1
USA_US101-15_1_T-1
USA_US101-15_3_T-1
USA_US101-16_1_T-1
USA_US101-17_1_T-1
USA_US101-18_1_T-1
USA_US101-20_1_T-1
USA_US101-21_1_T-1
USA_US101-2_1_T-1
USA_US101-22_1_T-1
USA_US101-22_4_T-1
USA_US101-2_2_S-1
USA_US101-2_3_T-1
USA_US101-24_1_T-1
USA_US101-25_1_T-1
USA_US101-27_1_T-1
USA_US101-28_1_T-1
USA_US101-29_1_T-1
USA_US101-3_1_T-1
USA_US101-3_2_S-1
USA_US101-5_1_T-1
USA_US101-5_2_T-1
USA_US101-6_1_T-1
USA_US101-8_1_T-1
USA_US101-8_5_T-1
USA_US101-9_1_T-1
USA_US101-9_4_T-1
ZAM_Ramp-1_1-T-1
ZAM_Tjunction-1_258_T-1
ZAM_Tjunction-1_263_T-1
ZAM_Zip-1_16_T-1
ZAM_Zip-1_20_T-1
ZAM_Zip-1_23_T-1
ZAM_Zip-1_26_T-1
ZAM_Zip-1_27_T-1
ZAM_Zip-1_34_T-1
ZAM_Zip-1_36_T-1
ZAM_Zip-1_50_T-1
ZAM_Zip-1_67_T-1
ZAM_Zip-1_7_T-1```Thanks for the feedback.
The USA_Lanker/Peach/US101 scenarios were hand-crafted in the early state of CommonRoad, thus I am not sure whether there is a feasible solution. I see that the USA_Lanker-1_10_T-1 scenario was solved during last year’s exercise, as we didn’t change the scenario data (merely upgraded to 2020 format), it should still be solvable. Maybe you can run on a different set of primitives and see if your planner gives you different solutions (which is already the case as reported by some students). Most of the scenarios on your list are created with a traffic simulator, so in theory, they should have a solution. I will inform my colleague who created the scenarios to have a look.
1 Like
I can’t solve the USA_Lanker-1_10_T-1 scenario with the three BMW_320i primitives, but it helps with the other scenarios I couldn’t solve, which does suggest that the motion primitive was the problem. Can we use multiple different motion primitives to solve the scenarios or is it necessary to solve all with the same motion primitive?
unfortunately, the current batch processing script only loads a default automaton for all scenarios. It would require some changes to be done in the batch_processing_parallel.py. As it is not easy to change and distribute it to others at this point, I would say it will be great if you can modify it so that it loads different automaton for different scenarios, and submit this file as well. We will incorporate this feature next time.
Can you also give us a hint how many (roughly) originally unsolvable scenarios are now solvable with switching the primitives? if there are still many unsolvable, then my colleague would have to examine the scenarios.
Out of the 164 were no solution was found I can now solve over 100.
1 Like
Just to make sure this is conveyed, you can submit multiple .yaml files instead of modifying the .py file for different automata: Submission of the CommonRoad exercise on Moodle
Hi, I am facing same issue as around same number of scenarios seems to have no collision free path to the goal area.
Where and how could we change the primitives or automaton to solve this? And what are the parameters that we cannot change, Like the car etc. ?
Great thanks
You can change the primitive in the batch processing config yaml file.
As far as I know you are not allowed to change vehicle and the cost function, not sure about the other parameters.
1 Like
Do you mean changing the time out and max_tree_depth? I thought you were talking about motion primitives and I thought the only way to change them was to change the vehicle-type or model, which I think was not allowed.
Or did I missed anything in the config.yaml ? 