Isaac Sim Position Randomization

Omniverse Isaac Sim
,
1

Isaac Sim Version

4.0.0

Operating System

Ubuntu 20.04

Topic Description

Hello,
I’ve been working with issac sim for a while trying to randomize the positions of a few carter robots. I’ve been trying to use the isaac sim replicator as outlined here:

https://docs.omniverse.nvidia.com/py/isaacsim/source/extensions/omni.replicator.isaac/docs/index.html

However, instead of setting the position to a distribution I set it to a random position in the costmap. I’ve been having a lot of problems with the robots having very weird behavior. I’m tracking the position of the robots in Rviz along with Isaac Sim. Sometimes the robots in Rviz will be on a different place on the map than the one in Isaac Sim. I notice that the robots seem to end up in a different position then what I pass into the position part of the replicator.

Does anyone have any experience with this? I’m not sure if I’m just using the replicator wrong or if there something else wrong with my code. Also, is there a better way to randomize the robots?

def reset_pos(randomization_node):
    with dr.trigger.on_rl_frame(num_envs=num_envs):
        with dr.gate.on_env_reset():
            #    dr.physics_view.randomize_rigid_prim_view(
            #        view_name=object_view1.name, 
            #        operation="additive", 
            #        position=rep.distribution.normal((0.0, 0.0, 0.0), (5.0, 5.0, 0.0)),
            #        velocity=[0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
            #    )
                dr.physics_view.randomize_articulation_view(
                view_name=carter_view.name,
                operation="additive",
                # joint_positions=rep.distribution.uniform(tuple([-0.5] * num_dof), tuple([0.5] * num_dof)),
                position = random_pos(robots[0],randomization_node),
            )
                dr.physics_view.randomize_articulation_view(
                view_name=carter_view2.name,
                operation="additive",
                # joint_positions=rep.distribution.uniform(tuple([-0.5] * num_dof2), tuple([0.5] * num_dof2)),
                position= random_pos(robots[1],randomization_node),
            )
                dr.physics_view.randomize_articulation_view(
                view_name=carter_view3.name,
                operation="additive",
                # joint_positions=rep.distribution.uniform(tuple([-0.5] * num_dof3), tuple([0.5] * num_dof3)),
                position=random_pos(robots[2],randomization_node),
                
            ) 

def random_pos(bot,randomization_node):
    """
    Generates a random valid position within the costmap.
    """
    # Costmap properties
    rclpy.spin_once(randomization_node, timeout_sec=0.001)
    # time.sleep(5.0)
    costmap = randomization_node.get_costmap(bot)
    print("checking costmaps")
    if costmap == None:
        print("No Costmap :(")
    if costmap:
        width = costmap.info.width
        height = costmap.info.height
        resolution = costmap.info.resolution
        origin = costmap.info.origin
        
        # print(f"width: {width}")
        # print(f"height: {height}")
        # print(f"resolution: {resolution}")
        # print(f"origin: {origin}")
        
        
        # Generate random coordinates in the grid
        x_index = random.randint(0, (width - 1))
        y_index = random.randint(0, (height - 1))
        
        # Convert grid coordinates to real-world position
        x = origin.position.x + x_index * resolution
        y = origin.position.y + y_index * resolution
        # x = -x_index * resolution
        # y = -y_index * resolution
        
        # Check costmap data to ensure it's valid (free space)
        costmap_index = y_index * width + x_index
    
        if costmap.data[costmap_index] == 0 and x > 0 and x < 10:  # Assuming 0 is free space
            print(f"Costmap Index: {costmap_index}")
            print(f"x: {x},y: {y}")
            return x ,0, 0
           
            # return x, y, 0
            # return [x,y]  # Return the random position as (x, y)
        
        return random_pos(bot,randomization_node)
   
    # return [random.randint(0, 50) ,random.randint(0, 50)]
    rep.distribution.normal((0.0, 0.0, 0.0), (0.0, 0.0, 0.0))

image
image1565×969 160 KB

Thanks!

2

@eminor2001 , is the position of the robot inside Isaac Sim different than what is beign shown in RViz on doing domain randomization with replicator?

3

Hello,
I made another post about this where I’m using function to randomize and I’m getting the same result. But yes it is on doing the domain randomization with the replicator.
Thanks!