UVDAR sometimes fails to detect

[lpmrz]

Hello,

I am doing formation control using relative distances. UVDAR works great when finding relative distances, but sometimes even in close range it fails to detect.

I exported four drones and each of them except the leader (leader moves freely, followers have to maintain formation) have two cameras on both sides and one on the back. I want zero communication between drones, that is why all followers have cameras.

I have 16gb of ram. Can this issue be related to lack of memory? Below is the status screen at the first few seconds of the simulation, as you can see third drone sees only two of other drones. Also Gazebo screen and my Htop screen are below.

Thank you everyone.

[klaxalk]

@ViktorWalter pls look into this.

[klaxalk]

@lpmrz could you please elaborate on the symptoms? Does it fail to detect completely? Or are there just short outages? Outages could be expected, since occlusions can happen. The simulation is rather realistic, so do not expect a 100% signal at all times.

[matemat13]

There seems to be plentiful RAM left, so that shouldn't be a problem. It's more likely as klaxalk suggested - maybe due to an occlusion or the targets simply being too far for reliable detection. Some false negatives are always to be expected in a real-world situation (with basically any detector) and because the simulation is quite high-fidelity, these occur also in simulated experiments, so your algorithms should be robust to these situations.

[matemat13]

closing due to inactivity

[ViktorWalter]

I'd need to see the output log from the uvdar observer tab (I'm assuming you are using sim_three_sided.launch). There are many possible explanations. It's even possible that one of the two virtual cameras of UAV3 failed to start.