With the BlueRobotics team significantly lowering the cost of admission for marine robotic applications and 3DR manufacturing the highly successful autopilot, Pixhawk and Pixhawk 2, I'm of the opinion it might be time to do some research into an affordable AUV that is more than a "one-off" and doesn't cost $50,000.
I'm in the process of finishing up my OpenROV Mid-Sized Observation Class ROV and turning my sights toward other projects to meet requirements in the marine sector. Rusty and I have had intensive discussions on the future of area, and its very bright. Although I'm starting to gather parts for an ASV project, an AUV is definitely on my list for something that I can drop off the side of a boat and quickly run a transect scan on a designated target and return with photos of the seabed.
Looking into the problem, the biggest hurdle is a vessel knowing its position underwater, or at least a decent estimation to guide it. From reviewing the wiki on the APM Extended Kalman Filter, it looks like this would fit the bill in "plane" mode. The EKF takes measurements from all available sensors and can work out position and velocity, even without GPS for short periods. Obviously, GPS corrections will be required periodically to prevent the propagation of error from becoming too large.
Although we can use GPS, the INS and magnetic compass, possibly the optical flow sensor, we would still need a depth sensor, sounder and associated interfaces to Pixhawk. I'm not quite sure how to develop and interface those. Maybe substitute the depth sensor for the barometer? And an acoustic sounder for the LiDAR?
In the very least, I am confident I can get a hull put together using the BlueRobotics 4" enclosure at maybe 36" long, and then 3D print a bow and tail section with a GPS and communications mast. I was looking at using either the MARES or COTSBot for the basic design layout. I favor using thrusters over control planes because 1) we don't have a reliable deep actuator 2) the less moving parts, the better.
Just some thoughts, as I'm not very far along designing this one yet, but I thought others might be interested or able to give their input. This intended to be a simple design that I place over the side, get it's bearings, dive, run a straight line about 10ft off the bottom, surface, get a fix, turn around and dive again to "mow the lawn". I've been working with others on imaging the seabed with Structure from Motion software to make up 3D models.
Replies
@Tom,
Thanks! It's a little better than paint, and I'm glad the BR team made their models available...so much easier to visualize things.
I have 3 of the T-200's with BlueESC's from the original KickStarter, so I'll probably just order one more and call it a day.
@Rusty,
Wow, I never thought of doing it like that! You are correct in that I am intending for this vehicle to have somewhat constant ahead motion. I would be concerned about it keeping an even keel for recording imagery and for the optical flow to register properly. I'd still like to see someone try it if they don't need to stay level.
Another option I am considering with the 4" Dome being released is something like this:
Anyone find that practical to have a viewport up front? I also haven't figured out a good mounting solution for the side mounted vertical thrusters to the 4" WTC.
Considering the batteries, I was thinking of something like these: LiFePO4 3.2V Prismatic Modules
Obviously I would need to tie more than one together in seines to get around 16V, but I like the even discharge rate, stability and safety. At least they will fit in the cylinder, on a 3D printed tray hopefully.
@Tom,
Thanks for the suggestion, that's what I was thinking of using. I would have plenty of 'wet-space' room in the aft tail cone to mount an encased power and signal board. They simplify the connections, but there will still be many wires back there I'll have to keep neat.
Speaking of control signals, looking through the APMPlane set-up and parameters, I can't do skid steering right from the code. Below are the outputs from the Pixhawk:
I'm thinking I can disregard the Aileron channel and install a manual mixer from the outputs of the Throttle and Rudder. They have stock V-tail mixers out there, but there are also some products from M-tronics out there:M-tronics Tank Mixer
@Tom
I thought APMRover didn't have a Z axis involved? Nor the EKF. Unless I could somehow move the Rover skid steering code over to plane, but I don't think that's happening
@Darius,
The EKF algorithm is already in the APM stacks. What is the UKF algorithm?
@Tom,
That was what I was thinking APMPlane might be for. It doesn't necessarily have to use the Rover code. Honestly, the vehicle is going to behave like a plane anyway, just in reverse.
I know I'm gonna lose the GPS, but the fun, technical side will be trying to keep the circle of error propagation as small as possible by using the extra sensors. I'm curious to see how well the optical flow sensor will do in shallow-ish water.
The depth will have to be interfaced with the Pixhawk or other controller somehow. Position and Depth will have to feed directly into the autopilot so it knows when to dive or surface. There is a Depth I2C module out there and we can use an external barometer...so I haven't given up yet.
Could you tell me why do you prefer to implement EKF algorithm over UKF algorithm ?
Spent the better part of today working on designs in 3DS Max. I'm still pretty terrible in AutoCAD, but I'm working through the tutorials when I can.
So from the first picture, you can see that my original idea of putting the vertical thrusters inside the nose and tail cones won't work. Already the thruster shroud pokes through a bit of the mesh on either side.
So on to the next revision. For this, I simply placed the thrusters on each end. I'm wondering how rigid this will be.
Below is the 36" version.
Below this is the 24" version.
I'm not going for super maneuverability on this one as it's mostly going to be used for long transects, but I'm thinking the 24" might be a better start and expand into the 36" if need be. Thoughts?
Also, any recommendations between the T-100/200 for something this size and power? I'll get into what I'm thinking for batteries in my next post.
Kevin,
Looking good! Have you considered eliminating the front thruster? The back thruster alone will provide you with pitch control much like an airplane elevator. You would no longer be able to descend directly downwards, but I assume that this vehicle will usually have forward motion and could dive and ascend like an underwater glider. It might be necessary to add fixed fins on the side (near the CG) with this method.
During descent, you would maintain forward thrust and then apply a small amount of upwards thrust to the rear thruster to point the nose downwards and descend.
Thoughts?
-Rusty
Sorry Rusty, my descriptions are garbage sometimes. Below is how things would get sandwiched together:
The nose/tail cones would bolt on over the top of the endcaps using recesses (not pictured) in the cone itself.
The printed pieces are not designed to be pressure resitant, only the enclosure. I would fill as much as possible the inside to add some rigidity, but other than that it would be free flooding.
Looking dead on, the nose could probably go all the way around, but the tail would have to have a section missing to access some of the penetrator locations for the vent valve, on/off switch, and a charge port so I don't have to take the thing apart each time.
Kevin,
Okay, great! I think this is simplest and best route to go. It'll be nice that you can have penetrations on both ends of the enclosure. Maybe you can find a way to quickly remove the tail cone so that you don't have to leave exposed access points for the vent and such.
-Rusty
@Thomas,
Thanks for your input, having an expert like you coming in is definitely appreciated.
I was gearing this first one more towards search and survey rather than surveillance. All the data and imagery it collects will have to be post-processed after the vehicle is retrieved. I'm quite sure others can come up with more advanced payloads, but my programming skills are pretty lacking.
Yes, I was planning on using 3D printed rounded extensions to the bow and stern with holes for the vertical thrusters and integrate side mounts for the horizontal ones. I'll make up a drawing here soon. Where I could, I was going to fill it with marine epoxy to make it more rigid.
The one design challenge I am going to have is I haven't decided how I want to mount the nose and tail sections to the endcaps. I think I have one of two options:
1) Mount it to the bolts that secure the end flanges
2) Create a bit of an extension to the cones and band them to the cylinder
I really like your ROV and I think it will do great, but I lack the heavy machining tools you have to cut ABS or HDPE. I went with the Makerbeam (6063-T5 aluminum) because it was pre-cut and easy for me to assemble and make modifications to. I also wanted to make something repeatable for someone else instead of a "one-off". Back when it only had 3 thrusters and was fully functional, I had it in the open ocean for about an hour, rinsed it off and had no corrosion or pitting either on the frame, brackets or bolts. I'm pretty happy with it so far. It should be finished by the end of the month.
@Darius,
I'm not up on the current Lego offerings these days, but if you have a control system in mind, go for it. I picked the OpenROV control for my ROV because it met my requirements for direct control with an HD video feed. I'm choosing to go with Pixhawk for the AUV because it handles autonomous control better. It's all a matter of control, not necessarily motors or pressure housings, I've seen pretty much everything used.
-
1
-
2
of 2 Next