I’m working on a 10ft autonomous offshore boat. I have some questions regarding obstacle avoidance and GPS.
1)Looking for a obstacle avoidance sensor and a sonar(lidar, radar, Proximity sensor,range finder and depth camera) to be used on the boat with Auto mode(Bendy Ruler) without affecting water as a obstacle.
2)I am using here3 with here+ base for rtk as the gps now. Looking for a rtk and gps baseline(both RTK and GPS for YAW). Can I use this gps for baseline(gps for yaw) and also use the same gps with this base station for rtk?
Thank you.
Hi @shan,
Object avoidance on boats is tough (actually it’s tough on all vehicles) due to the issue of false positives when lidar see the ground (or water in this case). It’s possible that a radar might work better because it should bounce off the water instead of “seeing” it. Nobody that I know of has actually tried this yet.
I’ve come to the conclusion that for good robust avoidance that actually adds value a 3D AI camera is required. We put some work into that during a GSoC project a couple of years ago but sadly it didn’t result in an easy-to-install and use solution. It might be a good basis for continued development but I don’t think we can expect it to just work.
Re the GPSs, I would have though that just using the H-RTK Unicore UM982 (Dual Antenna) would have been sufficient.
Hi @rmackay9,
Thank you for the response.I’ll try radar or 3d AI camera for obstacle avoidance. For the H-RTK Unicore UM982 (Dual Antenna) to have RTK, I need a separate base station/ ntrip connection. Or else what would you recommend as the base station if it’s needed?
Thank you,
Shan
Hi @shan,
Ah, right, of course. I didn’t read your comment about the RTK Base station.
I’m not an expert on the holybro product lineup but it does say on the Unicore GPS page that either this or this will work as a base station.
Hi @shan,
I worked on the mentioned GSoC project. With the latest update of the model generator, you can attach a cam to the boat, take some images (the images should be clean, like the demo video in the post: no high splash, no significant boat parts) and generate your own segmentation model following the instruction of a GUI. It’s not out of the box but it might be the best solution due to the fact that we are lack of both datasets and AI camera’s computing capability. If one day we can put a RTX40xx level card onto the boat, there won’t be need to distillate meta’s SegmentAnythingModel anymore. 
Also, the obstacle detection part is a bit rough (simply based on detection grids), but when you set the grid num to the same as pixel num, it actually outputs obstacle-only point cloud to ArduPilot. Then traditional SLAM, etc. can be introduced maybe.
And indeed, you may need some tweaks like adjusting the confidence thresthold before it works. It’s sad, but it’s true. 
Chenghao
1 Like
I’ll go with these GPSs. Thank you.