Tuesday, September 22, 2020

Foggy False Obstacles

This past week has been mostly filled with preparations for my upcoming pitch in the Ocean Startup Challenge this Thursday (the 24th), but I did find a bit of time to get out for a quick test on Monday morning. The air temperature was quite cold that morning (a low of 0 °C the night before) so there was a lot of fog on the water when AMOS was starting out:

Similar to what happened a couple of years ago on a foggy day (although with a different LiDAR model) AMOS was detecting a number of false obstacles due to the laser light from the LiDAR reflecting back from the water vapour. AMOS would turn in place until it found a "fog-free" direction and then proceed for a while before changing direction again. This slowed things down a bit at first, but soon the sun burned the fog away and AMOS sped from one end of Kelly's Creek to the other, a distance of almost 6 km in about 80 minutes:

I had to work pretty hard to tow AMOS back to the van in order to get back home in time for lunch and work at Measurand.

My slides and presentation are ready for the pitch on Thursday, although I'm meeting with some people tomorrow to get their feedback, so they might still need a bit more modification. Anyone who is interested can view the slides and speaking notes here:


Tuesday, September 15, 2020

More Capacitors = More Speed

 Work this week continued on efforts to re-wire the electronic speed controller (ESC) behind the propeller motor. This took way longer than I had planned, and involved swapping out the motor, swapping out the electronic speed controller, re-soldering all of the connections, re-calibrating the ESC, learning what all the ESC beeps meant and modifying the calibration software to be able to adjust all of the parameters according to those beeps. No matter what I tried though, the motor was always very choppy when run at medium to high speed. With nothing left to try, I resorted to Google, which fortunately came up with this web site that had a good discussion about some of the problems that can result with an ESC if your cables delivering DC power from the battery are too long: https://www.rcgroups.com/forums/showthread.php?952523-too-long-battery-wires-will-kill-ESC-over-time-precautions-solutions-workarounds

By positioning the ESC outside the box and behind the propeller, I had lengthened the cables delivering the 12 VDC power to the ESC by about a foot or so. The problem is not the extra bit of resistance that this adds, but rather the extra inductance that it adds. The inductance results in voltage spikes at the ESC, which can apparently result in erratic performance of the motor. The solution is to add extra capacitance at the ESC to help smooth out the voltage fluctuations. There is already a 30V 220 μF capacitor on the ESC, but the rule of thumb specified in the above website is to add an additional 220 μF capacitor for each additional 10 cm of power wire length. I added three 330 μF capacitors in parallel over the voltage inputs to the ESC. They were only rated for 25 V, but they seemed to improve things quite a bit, and were able to allow the motor to run at a normal speed.  I also tried adding an additional 330 μF capacitor to see if it might allow the speed to increase further, and it did a little bit, although not quite as much as I had hoped. 

So now there is an ugly cluster of capacitors and an ESC mounted behind the propeller. These would need to be weatherproofed with some silicone or other material if used outdoors on a consistent basis. Better positioning of the solar charge controller would also allow me to shorten the wire leads that power the ESC. Really the only reason for positioning the ESC behind the propeller is to keep it cool. Having it inside the electronics enclosure risks having it heat up to 110 °C on hot days, at which point the ESC firmware immediately cuts its output power by 50%. This sort of thing was happening occasionally back in August when the weather was sunny and the outdoor temperature was close to 30 °C.

Cluster of capacitors are shown here circled in red.

On the weekend I took AMOS and its new capacitors downtown for a test run, and was happy that it beat its old record for going up and down the downtown section of river by 5 minutes: from 47 minutes to 42 minutes. The dissolved O2 levels were good too - I guess those are going back up again now that the water is cooling again:

Tuesday, September 8, 2020

Ocean Startup Challenge!

 There was a big announcement today for In Nature Robotics: we were one of 31 companies out of 158 applicants to be picked for the next round of the Ocean Startup Challenge: http://www.oceanstartupchallenge.ca/announcements/. Starting this Friday, there will be a week of presentations and workshops to get ready for a final pitch presentation, sometime between September 22 and September 24. The 31 companies will be competing for 10 prizes of $25k plus in-kind support. It seems sort of similar to what we went through back in May for the Volta competition; except more of an overall ocean theme. 

I took a couple of days off from AMOS this week to go on a fun family camping trip at Spednic Lake. It would have been nice to try out AMOS there, but the van was packed tightly to the roof with camping supplies, so it wasn't possible. I did manage to get some 3D models of the new WeatherBox put together and tested for water-tightness though. Here is a picture of the component parts (no more little nuts or bolts required!!!) and a picture of the put-together unit (minus the interior locking piece):

The only thing that needs to be modified is that the camera board is recessed a bit too far inside the enclosure, so that you can see tiny smudges of black (from the enclosure) in the corners of the field of view:

(Please ignore the terrible state of this lawn. 😀)

To fix this, I can try to either (i) elevate the level of the camera board inside the enclosure, or (ii) trim down the thickness of the enclosure a bit. 

I have also been working on re-wiring the electronic speed controller (ESC) to situate it behind the propeller. Unfortunately though, I think I may have broken one or more of the wires going into either the ESC or the propeller motor today, as it was functioning sporadically, depending on how I moved the wire around. I'll try a replacement motor tomorrow, and if that doesn't work I'll try replacing the ESC.

Tuesday, September 1, 2020

Back From The Wilderness

 Avid readers of the blog may have noticed that there was no entry last week; that was the first time in over 2.5 years that the weekly update was missed! Without going into too many details, a long 15 hour bike ride through New Brunswick trails and back roads turned into an even longer 24+ bike ride due to some navigation errors that resulted from poor preparation (forgot 1 phone charger and the other 2 chargers that were brought were not charged, did not bring a printed map (only printed directions), and did not bring proper equipment to change a bike tire, which turned out to be necessary for the last 50 km or so). After being rescued and returned to safety, I was in no condition to think about the blog, and thereby felt justified in skipping a week to catch up on some sleep. Many thanks to Kelly and Dad for searching for us and rescuing us after this misadventure. 

The last time I had AMOS out for a test run was on August 22 at Kelly's Creek. One thing I happened to notice around the middle of this test was that the air propeller motor would sometimes audibly switch down to a lower speed, resulting in a somewhat lesser thrust. Going with the wind the reduction in thrust was not an issue, but on the return trip it was. Some testing in the backyard the next day revealed that the electronic speed controller (ESC) board output power became significantly less when it overheated. Currently, the ESC is sitting in the bottom of the back electronics enclosure, with the battery and a bunch of other electronics, so it makes sense that it could get pretty hot, especially if the weather outside is sunny and warm. By positioning the ESC outside the box, behind the propeller in the airflow however, the output power remained stable, presumably because the ESC was sufficiently cooled. Apparently drone makers know this already; there are instructions available on the Internet for applying a silicone conformal coating to the ESC board to protect it against moisture. The weather is starting to get cooler here, so I might not notice the overheating effect in future tests, but I think I'll re-wire things to put the board outside behind the propeller anyway. 

 Efforts are currently underway to re-design the WeatherBox (i.e. the waterproof enclosure for the camera on AMOS). I would like to get rid of the nuts and bolts and just have a window cap that screws directly onto the enclosure. This would make assembly a whole lot easier, and should look better too. Here are 3D printed versions of the base enclosure piece and an initial window cap (without the plexiglass window) screwed together:

This past week was Jata's last as an intern at In Nature Robotics. One of her projects was an online 3D modeling / ordering tool for AMOS. I have been trying to teach myself 3D modeling using TinkerCAD and other 3D tools to continue her work, but so far have not had much luck. Importing and exporting with these tools tends to do unexpected things to the models. Possibly it is just user error though!

Tuesday, August 18, 2020

Read The Fine Manual

 Last Wednesday I was able to confirm by looking at the end of the dissolved O2 probe with a magnifying glass that there was a small crease in the membrane that was allowing air and water to leak in and electrolyte to leak out. So that confirmed why the previous readings had been getting progressively worse. Here is some better data collected at Woolastook with more electrolyte added and a new membrane over the end of the probe:

As you can see in the above screenshot, this time AMOS followed a route through the middle of the water, where it was relatively deep and there was little danger of damaging the end of the probe by dragging it against rocks, sticks, or other objects. The following graphs show the dissolved O2 data collected so far out at Kelly's Creek and near downtown Fredericton:

The results recorded in early August should be ignored, as the probe was most likely leaking at that time.

We entered an application in to the Ocean Startup Challenge last week: https://www.oceanstartupchallenge.ca/. I know of at least a few other people that have also entered submissions, so probably this contest was quite popular. On September 04, we will know whether or not our application made the short list. 

Just this evening I finished a 6-page quick start manual for getting up and running with AMOS. It hasn't been tested on anyone yet, although Kirsten volunteered to try it out and see if she could get it to work, so I'll try to take her up on that. Check out our support page: https://www.innaturerobotics.com/support to have a look. RTFM! 😀

Tuesday, August 11, 2020

Leaky Dissolved O2 Sensor?

 After acquiring a few weeks of experience with the new dissolved oxygen sensor on AMOS, I am unfortunately a bit disappointed with its performance. Software-wise and electrically it worked great. They have some nice simple I2C and serial commands for getting data and calibrating, and you can get temperature compensated data at a rate of about 1 Hz. The bad part is that the calibration doesn't seem to hold for more than a few hours. The sensor probe is a galvanic one, in the sense that it uses an anode and cathode surrounded by an electrolyte to produce an electric current (and small voltage) when oxygen molecules cross over the probe's membrane. If I fill the probe with electrolyte (requires ~ 2 ml), calibrate, and then do a test run with AMOS, the results seem generally pretty good. If I go out again even just a few hours later, the results seem a bit lower, and then lower still the next day. Even re-calibrating doesn't really seem to work very well if it has been a long time since the initial filling of electrolyte solution. The membrane on the probe is quite thin, and a bit flimsy (slightly thicker than saran wrap), so I'm wondering if it just doesn't make a very tight seal? The online shop where I bought the sensor sent me an email request to review it, so I gave it a mediocre 3 stars with a description of the issues I was having. They said that they would forward my comments to the manufacturer, so we'll see if I hear back. 

Anyway, here are some pretty good results obtained near downtown Fredericton on Saturday, about an hour after filling with electrolyte and calibrating:

The dissolved O2 readings were all within a narrow band of 9.1 to 9.2 mg/L, which is a normal, expected level for the amount of oxygen dissolved in water. 

Later that afternoon I took AMOS out to Woolastook to collect some more data, but this time the readings started off a bit under 9 mg/L, jumped around a bit, and then seemed to gradually drift lower to under 5 mg/L as AMOS traveled along the 4 km route:

Today I re-visited Woolastook, and performed a re-calibration before leaving, but did not re-fill any of the electrolyte. This time the dissolved O2 started off at around 7 mg/L, and then dropped down to nearly 0 mg/L about halfway through the test: 

Since I saw a number of small fish swimming around, I don't think the level could have possibly been that low! I'll try going back tomorrow, but will top up the electrolyte again to see what difference that makes. The probe manual is vague about how often re-calibration is required, but does say that the electrolyte should last for about 2 years before it is depleted. Perhaps there is an issue with dragging the probe behind AMOS? It is mostly horizontal while AMOS is moving. Or maybe the probe membrane is getting damaged by dragging it through shallow water, grass, etc. close to the shoreline? I'll try not to waste too much more time on this, but these things kind of bother me! 😖

Tuesday, August 4, 2020

Paddle Test Take 2

Despite the 3-D printed "paddle fail" of a couple of weeks ago, I had not given up on the idea of 3-D printing something for paddle propulsion. This time around, I made a simple rectangular paddle with a hollow tube along one edge for mounting onto the motor shaft. To test out the concept and see if a pair of paddles would generate sufficient thrust for AMOS, I used a chisel to gouge out some motor mounting holes in the old surfboard, and wired up (this time properly!) the DC motor controller to an Arduino Uno, and wrote a simple program for driving both motors at the same speed. 

It seemed like maybe it was faster than the air propeller version, although to be certain I guess I would need to program both boats for the same GPS course and race them against each other. It still remains to be seen how the paddles would fare in some of the weedy locations that the airboat frequents. This video recorded on the weekend shows the airboat version of AMOS moving through some pretty dense river grass (although the depth transducer and turbidity probe had to be pulled out of the water to minimize drag):

While AMOS was traveling through the river grass, it was also measuring the dissolved oxygen content of the water. From the few tests that I have done in the river near downtown Fredericton and out at Woolastook over the last couple of weeks, it looks like the dissolved oxygen content is dropping.  Here is the latest dissolved oxygen data from downtown Fredericton:

Dissolved oxygen content near downtown Fredericton on Aug. 02, 2020 varied from 5 mg/L (purple) to 9.1 mg/L (red). 

Unfortunately our humid weather over the last few weeks coupled with repeated testing has dissolved the construction adhesive that was bonding the foam pontoons to the aluminum plates. To remedy this, I spent about $13 for a tiny package of two-part "marine" epoxy. It is supposed to offer a bonding strength of 4000 psi, and be waterproof. So we shall see. I re-bonded the delaminated sections a few days ago, and it seems to be holding quite strongly so far!