I have been researching the different versions of Obstacle Avoidance (OA) here on ArduPilot and a few other forums to zero in on cost effective reliable hardware.
Question One:
RPLidar A2 is noted on the configuration page (with sampling rates of 4k & 8k).
Slamtec has a few other options that have similar power and size dimensions with 32k sampling rates at a comparable or lower cost.
Has anyone used the RPLidar A2M8-R6 (lower res/cost unit) or the RPLidar S2 or S2L…both with 32k sample rate and better range?
Question Two:
Can Nvidia Jetson be integrated with ArduPilot?
Has anyone successfully used the Jetson onboard medium size drone?
We are using an ouster lidar and a jetson tx2 running ROS. We send the lidar messages over mavlink. OA worked when we were using 4.1.1 firmware, but since we have updated to 4.2.3, OA has stopped working. Only avoid is working in manual modes. I posted about this a couple of days ago in the 4.2 forum with no reply.
I have been researching since my OP…and have started focusing on a solution based on either:
Four of the Benewakes TF02 pro (four sides with slight angle down)
or
three of the stereo vision IMX219 w/160 degree FOV. (front, back and bottom)
I am leaning toward the IMX219 as it also has the video feed for facial recognition, FPV and tracking/following functionality.
Still working on getting multiple devices connected simultaneously to the Nano (i did read about a version of the Nano that includes more than one camera input…but the three i have are all B02 version which just has the single camera connect).
I did consider the Realsense d430 (well, the OEM Dell laptop version as you can pick those up for $60 + $20 for the USB power conversion) instead of the $200+ cost of the retail d430
With my implementation, I am afraid, yes you need all the computing power you can get. The TX2 already runs at its limits.
But there are a lot of new solutions out there. You may look at the new ZED2i where a lot of the 3D imaging processing has been moved into the camera.