[Feature] Allow customized, fixed viewer bound in interactive mode

[Jianye-Xu]

Current, VMAS seemly does not support a customized, fixed viewer bound in interactive mode. This feature would be useful if the origin $(0,0)$ of a user's map is not centered. In my case, the origin of my map is located at the lower left corner, meaning that the $x$-range of my map is [0, world_x_dim], and the $y$-range of my map is [0, world_y_dim], instead of [-world_x_dim/2, world_x_dim/2] and [-world_y_dim/2, world_y_dim/2]. In addition, I do not want to shift my map since I am using an HD map with dense data.

My current workaround is to add an attribute viewer_bound in my Scenario class's make_world function:

self.viewer_bound = torch.tensor(
    [0, world_x_dim, 0, world_y_dim],
    device=device,
    dtype=torch.float32,
)

and then modify the source code of file /vmas/simulator/environment/environment.py around this line from

self.viewer.set_bounds(
    -cam_range[X],
    cam_range[X],
    -cam_range[Y],
    cam_range[Y],
)

to

if hasattr(self.scenario, "viewer_bound"):
    self.viewer.set_bounds(
        self.scenario.viewer_bound[0], # left boundary
        self.scenario.viewer_bound[1], # right boundary
        self.scenario.viewer_bound[2], # bottom boundary
        self.scenario.viewer_bound[3], # top boundary
    )
else:
    self.viewer.set_bounds(
        -cam_range[X],
        cam_range[X],
        -cam_range[Y],
        cam_range[Y],
    )

This way, if self.scenario contains the user-given attibute viewer_bound, the viewer bound will be set accordingly; otherwise, the funtionality is untouched.

I would appreciate it if anyone could point out if there is an easier way to achieve this. If not, I think it would be nice if the developer would add such a feature to VMAS.

[matteobettini]

Hey!

Thanks so much for the comment. I think I get your request.

In other words you want to have the center of the render image in a custom point and not in [0,0].

My suggestion is this:

• have an attiribute in the scenario called render_origin which defaults to [0,0] and is used only when agent_index_focus=None in render()
• this param would be added to the line you linked like so

self.viewer.set_bounds(
        -cam_range[X] + render_origin[X],
        cam_range[X]  + render_origin[X],
        -cam_range[Y] + render_origin[Y],
        cam_range[Y] + render_origin[Y],
    )

• this way you can override this and set it to [ world_x_dim/2, world_y_dim/2]
• in addition to your solution this would also zoom out in case any agent goes off the map (which you can easily prevent by setting the world bounds)

What do you think?

[matteobettini]

I am down to enable this feature.

With this added you would have control over:

• scenario.render_origin
• scenario.viewer_size
• scenario.zoom

Which should let you do any rendering you want I think

[Jianye-Xu]

Thanks for your suggestion!

I tried it and think it would be a good solution. There are two small issues I encountered:

• There are blank margins around my map. I tried to adjust scenario.viewer_size but it did not change anything.
• The viewer bound is dynamically changing, which is unexpected since I am using a closed map and agents will be reset once they across the map boundaries.

Do you have any ideas to fix these two issues?

This is a video demo, which hopefully could help in understanding:

viewer_center.mov

[matteobettini]

Wow that scenario looks so cool!

• for the first issue: viewer_size only determines the aspect ratio, if your map is a square i would not change that, have you tried changing the zoom?
• For the second issue: the bounds should dynamically change only if an agent gets within $\epsilon$ of getting out of sight. I thought this should not happen for a reasonable zoom, but we can have anotehr param scenario.fixed_zoom to make it fixed

[Jianye-Xu]

Thanks!

• Yes, I also tried to change scenario.viewer_zoom but it did not change anything.
• Yeah this is also want I expected. But somehow the bounds still changed dynamically when some agents moved, not to mention that the map was already zoomed out a lot.

My setups are as follows:

world_x_dim = 4.5  # [m]
world_y_dim = 4.0 # [m]

self.viewer_zoom = 1.0 # The adjustments of this parameter did not influence anything in my case
self.viewer_size = (world_x_dim * 200, world_y_dim * 200)

Do you have any clues?

[matteobettini]

Could you try out freshly in #93 please?

I should have fixed a few things

[Jianye-Xu]

Thank you! I tried it and there are no unwanted blank margins now. Most of the time, it works very well. However, when all agents are close to each other, the view zooms in significantly (see the Fig. below), which leads to the loss of the map's overview. This is beneficial as it highlights the agents rather than the map's overview. However, sometimes I think it would be nice to have a fixed viewer bound, for example, when I put the evaluation videos of different policies together to intuitively compare their performance, the dynamically changing zoom may be somehow distracting.

What do you think?

[matteobettini]

The dynamical zoom should not happen if you set scenario.viewer_zoom to a high enough value.

try incrementing that parameter up to a certain value and you should get what you are looking for

Otherwise let me know and i will investigate further

[Jianye-Xu]

Thanks for pointing this out. This is exactly what I am looking for!

I think you can close the issue now.

[matteobettini]

Very glad it worked!

From the rendering this scenario looks super exciting! Would love to add it to vmas if you feel like contributing it one day!

[Jianye-Xu]

Sure, I would definitely love to! I will let you know when the time comes!