Random destination generator

[AndreasKaratzas]

What feature are you suggesting?

Overview of request and why would this feature be useful:

I used AirSim to train a reinforcement learning agent. The purpose was to make the vehicle reach its target by training that agent. It was really difficult for me to solve overfitting of my model. A feature that could help AirSim users to solve any issues regarding similar issues would be a random destination generator method. This method would return a Vector3r instance of some random coordinates in a selected map. The only guarantee of the random destination generator method would be the feasibility of the destination, in other words there must be an accessible path between the vehicle in use and the generated coordinates. The only optional variable would be a distance (radius) threshold between the vehicle and the generated coordinates. That way we could provide such RL agents with a sufficient amount of random data they need to properly train.

Smaller Details:

def random_destination_generator(running_map: str, running_vehicle: VehicleClient, distance_from_vehicle: int)

Nature of Request:

Addition

[jonyMarino]

Hi @AndreasKaratzas! I think you can accomplish this by running a path planner in ROS. Take a look at this discussion to integrate with ROS navigation stack: #3275