Many problems not possible

CommonRoad-Search
1

Hello,

I am a little bit upset because I already spent multiple days on this project due to the fact that I thought the project is really cool, but some solutions are impossible to solve. I tried multiple hours to solve some ACC tasks, because I implemented an ACC control unit & a predictive collision warner from another lecture and I thought it would be easiest task for debugging. But in these examples:

[ DEU_A9-2_1_T-1]

[ DEU_A99-1_1_T-1]

[ DEU_A99-1_2_T-1]

[ ZAM_ACC-1_3_S-1]

[ ZAM_ACC-1_3_T-1]

[ ZAM_HW-1_1_S-1]

the v_init is bigger than the maximum speed of the vehicle as you can see here(from the trajectory file of the bmw (last trajectory with fastest v):

10

0.0
-0.0
<steering_angle>-0.0</steering_angle>
20.0
-0.0
<time_step>0</time_step>


10.0
-0.0
<steering_angle>-0.0</steering_angle>
20.0
-0.0
<time_step>5</time_step>



2.0
-0.0
<steering_angle>-0.0</steering_angle>
20.0
-0.0
<time_step>1</time_step>


4.0
-0.0
<steering_angle>-0.0</steering_angle>
20.0
-0.0
<time_step>2</time_step>


6.0
-0.0
<steering_angle>-0.0</steering_angle>
20.0
-0.0
<time_step>3</time_step>


8.0
-0.0
<steering_angle>-0.0</steering_angle>
20.0
-0.0
<time_step>4</time_step>



Or am I missing something?

Thank you.

H.

2

Hi,
Yes, it is true that some of the early scenarios might not be solvable (as the validity checker had only been added afterwards). However, in the list that you have provided, DEU_A9-2_1_T-1 is solvable, see here.

4

Hello vitalii_risinov,

thank you for the quick answer.
For some problems the main problem for a saftey orientated motion planner is the time, because the scenarios are often finished when the acc-target vehicle stops in the middle of the goal area.
If you use an alogrithmn where a saftey distance is used (x_0+v_ego*2) its really hard to successfully do them, because its avoiding small gaps where the car just fits in. Examples are:

[USA_Lanker-2_24_T-1]

[USA_Lanker-2_5_T-1]

[USA_Peach-3_2_T-1]

[DEU_Ffb-1_3_T-1]

Some more, don’t remember the names

Is there a way to increase the time horizont a little bit (2s) for the scenarios that saftey ortientated algrithmns with saftey margin can cover the missing distance?

Thank you and have a nice sunday.

H.

5

I don’t know. Let’s see if somebody answers. :slight_smile: