Problem on detecting goal state

1

Hi,
I’m using GBFS as search algorithm and I made some changes in the heuristic.
I’m facing the problem that the car successfully finds the road up until it reaches the final goal. When its close to the goal it happens that:

  • Sometimes it goes over the goal, and search continues on
  • Sometimes it slightly deviates the road map as soon as it approaches the goal.

Do you have any recommendation e.g what kind of attribute to take into consideration?

79347233_2520524871512442_3818018453453799424_n
79347233_2520524871512442_3818018453453799424_n1186×958 88.1 KB

2

Hi,

  1. It is normal that the trajectory goes over the goal (in terms of position), because we always concatenate the whole primitive. However, keep in mind that in the end we only return a pruned trajectory, meaning that states that are behind a state that reaches the goal will be discarded.

  2. What do you mean by deviating the road map? If you meant that it steers significantly near the goal, it is most probably caused by the weights of the heuristic terms. I.e., at some states near the goal region, the ones steering to the goal has lower costs, thus they were picked.

3

Hey @ge68wug,
have a look at the time_step of the first state that you expect to be within the goal.
If the time_step is below the minimum or above the maximum it will continue. Next I would look at the orientation and check that it’s within the specified interval.
These were at least the problem I encountered with the first algorithms.

4

In scenario Lohmar 57 the car reached the goal region but the time_desired is from 99-100, while the car arrived on time step 30.

The search algorithm just goes on till time out (default set at 30). What should I do there?

5

your car should slow down to reach the goal within the specified time interval : )

6

Sorry for the stupid question, but may I ask how I can let the car slow down XD?

So far I think we can only choose path based on the algorithm?

Best regards