srmainwaring
commented
1 year ago Hi @ConnorDTaylor, the gpu_lidar sensor registers returns from the wave surface when using the latest version of the wave plugin with Gazebo Garden.
In the example below I've added a lidar sensor to the waves.sdf example using the sdf
<model name="gpu_lidar">
<static>1</static>
<pose>0 0 1 0 0 0</pose>
<link name="base_link">
<visual name="visual">
<geometry>
<box>
<size>1 1 1</size>
</box>
</geometry>
</visual>
<sensor name='gpu_lidar' type='gpu_lidar'>
<topic>lidar</topic>
<update_rate>10</update_rate>
<lidar>
<scan>
<horizontal>
<samples>60</samples>
<resolution>1</resolution>
<min_angle>-3.14159265</min_angle>
<max_angle>3.14159265</max_angle>
</horizontal>
<vertical>
<samples>8</samples>
<resolution>1</resolution>
<min_angle>-0.261799</min_angle>
<max_angle>0.261799</max_angle>
</vertical>
</scan>
<range>
<min>0.08</min>
<max>10.0</max>
<resolution>0.01</resolution>
</range>
</lidar>
<alwaysOn>1</alwaysOn>
<visualize>true</visualize>
</sensor>
</link>
</model>
The waves are displayed as wireframe for clarity and only the rays that hit are shown. There isn't a currently a sonar sensor in gz-sensors, but you may be able to adapt the Lidar sensor output for your requirements.
I don't have a Gazebo classic environment set up at present, but it may be worth checking the gpu_lidar in that environment to see if you get similar behaviour.
ConnorDTaylor
This questions was asked to Rhys via email, here is the conversation history up to this point for anyone interested.
My initial email: I am trying to use your ASV wave simulator and I have some issues with the collision of the waves. I am trying to use sonar sensors to measure the wave height in Gazebo, but the waves being generated are purely visual and have no collision (which you know).
So, I am wondering: is there any way to apply collision to the visual wavefield? Or is there a way to apply the Gerster wave .vert and .frag files to some sort of collision mesh so I can have my sensors detect the water? I am working on Ubuntu 18.04 with ROS Melodic.
Any help you can provide would be greatly appreciated.
The first response from Rhys: The physical aspects of the waves are handled by the buoyancy and hydrodynamics physics plugins that are attached to the model. If the waves had a collision as well then the objects would sit on top of it (like a terrain), and there would be a significant performance impact as the collision algorithm would need to examine all the wave triangles.
The waves do return depth information when sensed by a depth camera, so perhaps another approach would be to look into how your sonar sensor is getting a return. Is it just the standard sensor available with Gazebo or something custom?
Second email to Rhys from me: I figured giving the waves' collision would cause the simulation to be too computationally intense (especially for my hardware I think), though my solution for this would be to make the wave area as small as possible.
As for my sensor, I am using the sonar sensor from the Hector Quadrotor package. I believe this is simply a ray that looks for something to collide with to return a height.
To solve this problem I was considering trying to adjust the sensor to detect points based on the Gerstner waves being generated by the Wavefield script. However, I am confused about what element of this file defines the calculated wavefield at each time step, could you provide any insight into this?
I am looking to pull the z components of all of the points in the wavefield so that I can use these as conditions for sensor detection.
Answer to the second email from Rhys: I think the approach you’re looking for may be similar to the way a GPU LIDAR sensor is implemented in Gazebo. I think this does record a return from the waves (even for the earlier version using the Gerstner waves scripts).
There is a simple wave example included in the more recent Gazebo distributions. I’ll take a look at whether that example results in a return from the waves for a start and from that we may come up with a way for you to estimate wave heights from a range sensor.
A method of last resort would be to use the same ray intersection calculation used for the wave height in the hydrodynamics plugin (but tracing the ray at a given angle from the sensor location). That would be less efficient than a GPU based approach using the scene depth buffer though.
The rest of this conversation will be had here so anyone can access the information. Thanks for the help Rhys!