why my vehicle don't move?

[lsm2842035890]

I loaded the gridsmall map and paved a simple road. I want the vehicle to run according to my trajectory. Why does the vehicle not move?

[ASionville]

For now I think your "AI" stops with the python script.
You should add time.sleep(x) at the end of script for it to be "alive" for x more seconds and actually move.
You can put it in a while loop for better control.
You can consider using bng.step(n) too if you want, n here is the number of steps to go ahead to if I remember correctly

[lsm2842035890]

I tried that, but I still can't get the vehicle to move according to my coordinates. Could it be related to the map? The map I loaded is gridsmall, and the road sign information in it is incomplete, making it impossible to move the vehicle.

[ASionville]

If you put the AI in span mode, does it move at all ?

[aivora-beamng]

Hi @lsm2842035890,
the script must be "realistic" (the vehicle AI tries to get the vehicle to the point (x, y, z) in the time t. You may use the ScriptAI Editor to create valid scripts. Also, the looped argument of the road_a road should be set to False, as your road is a straight one.

This code works for me, you can try yourself:

from beamngpy import BeamNGpy, Road, Scenario, Vehicle


def generate_scripts():
    script = [
        {'x': 0, 'y': 300, 'z': 0.1, 't': 0},
        {'x': 0, 'y': 400, 'z': 0.1, 't': 10},
        {'x': 0, 'y': 500, 'z': 0.1, 't': 20},
        {'x': 0, 'y': 600, 'z': 0.1, 't': 30}
    ]
    return script


def set_scripts():
    bng = BeamNGpy('localhost', 64256)
    bng.open()

    scenario = Scenario('smallgrid', 'custom_road_network')

    road_a = Road('track_editor_C_center', rid='center_road', looped=False)
    road_a.add_nodes(
        (0, 300, 0, 6),
        (0, 400, 0, 6),
        (0, 500, 0, 6),
        (0, 600, 0, 6)
    )
    scenario.add_road(road_a)

    vehicle = Vehicle('ego_vehicle', model='etk800')
    scenario.add_vehicle(vehicle, pos=(0, 300, 0), rot_quat=(0, 0, 1, 0))
    positions = [(0, 400, 0), (0, 500, 0), (0, 600, 0)]
    scales = [(1., 1., 1.)] * 3
    ids = ['my_waypoint_1', 'my_waypoint_2', 'my_waypoint_3']
    scenario.add_checkpoints(positions, scales, ids)
    scripts = generate_scripts()
    scenario.make(bng)
    bng.scenario.load(scenario)
    bng.scenario.start()

    vehicle.ai.set_speed(30, 'set')
    vehicle.ai.execute_script(scripts)


set_scripts()
input('Press Enter to exit...')

[lucena564]

I was with the same point. That function here:
start_recording()
stop_recording()
https://beamngpy.readthedocs.io/en/latest/beamngpy.html#beamngpy.api.vehicle.VehicleApi

It should record that points for us? Have some feacture that can helps with it? Or it is like try and error?

In ai_line.py we have points passed by sin function, like:

But if I load that map:

scenario = Scenario('west_coast_usa', 'Loop 1 da Rodovia')

It will load this:

I dont have the points to follow, but have someting to get the point at the moment to build it manualy?

Was thinking about run that using:

vehicle.ai.set_mode('disabled') 

[lucena564]

I got those points here:

script = [{'x': 460.86572265625, 'y': 328.2005615234375, 'z': 119.28526306152344, 't': 10}, {'x': 430.08746337890625, 'y': 301.0779724121094, 'z': 119.33838653564453, 't': 20.0}, {'x': 362.0964050292969, 'y': 250.7567138671875, 'z': 119.33403015136719, 't': 30}]
points = [(460.86572265625, 328.2005615234375, 119.28526306152344), (430.08746337890625, 301.0779724121094, 119.33838653564453), (362.0964050292969, 250.7567138671875, 119.33403015136719)]
sphere_coordinates = [(460.86572265625, 328.2005615234375, 119.28526306152344), (430.08746337890625, 301.0779724121094, 119.33838653564453), (362.0964050292969, 250.7567138671875, 119.33403015136719)]
sphere_radii = [0.0, 0.246201938253052, 0.08550503583141716]
sphere_colors = [(0.0, 0, 0, 0.8), (0.984807753012208, 0, 0, 0.8), (-0.34202014332566866, 0, 0, 0.8)]

scenario = Scenario('west_coast_usa', 'Loop 1 da Rodovia', description='Spanning the map with an advanced IMU sensor')
vehicle = Vehicle('ego_vehicle', model='etk800', license='RED', color='Red')
scenario.add_vehicle(vehicle, pos=(460, 328, 119.28), rot_quat=(0, 0, 0.3826834, 0.9238795))

But looks that AI is not looking collision, just going on to the left wall. Those points have time enough to go across.

[lucena564]

I dont know if it will help anyone but..

The problem was that I was using the function to get my positions of the car on the street:

vehicle.get_bbox()

It returns:

{'front_bottom_left': [460.85791015625, 328.1919860839844, 119.28421783447266], 'front_bottom_right': [462.30120849609375, 326.7507019042969, 119.31969451904297], 'front_top_left': [460.831298828125, 328.1995544433594, 120.67394256591797], 'front_top_right': [462.27459716796875, 326.7582702636719, 120.70941925048828], 'rear_bottom_left': [464.15704345703125, 331.4969177246094, 119.32947540283203], 'rear_bottom_right': [465.600341796875, 330.0556335449219, 119.36495208740234], 'rear_top_left': [464.13043212890625, 331.5044860839844, 120.71920013427734], 'rear_top_right': [465.57373046875, 330.0632019042969, 120.75467681884766]}

The point is that the grid of the ground is a little bit higher then the grid of the car, so, when I passed the cling=True that makes the car try follow a really deep point (the first ground below the road grid). Looks that:

It is possible observe that have a line that going deep in to the ground, it is the second point that the car should go to, was the point of the bug.

To fix that needed use an offset of 0.9 at my node in 'z' position.

node = {
            'x': (points['front_bottom_left'][0] + points['front_bottom_right'][0])/2,
            'y': (points['front_bottom_left'][1] + points['front_bottom_right'][1])/2,
            'z': points['front_bottom_left'][2], # + 0.9,
            't': (current- initialization).total_seconds()
        }

Using that we can get someting like that: