Friday, September 24, 2021

AMOS Captures Some Fish

 Things have been busy lately, and there hasn't really been much occasion for taking AMOS out in the field. Over the last week though, I was able to develop some software for saving 30 fps 1920 x 1080 video files on AMOS, while still continuing to do all of its regular navigation and data saving stuff. Testing and debugging was done in the pool, and everything seemed to be working well. 

This morning I took AMOS out to Kelly's Creek to test how the navigation, sensor data saving, and video recording all worked together. I was worried that the addition of the video recording in a real environment would slow down the other functions, possibly resulting in sluggish handling or missed data samples. Everything seemed to work pretty well though, although when I had plotted the course the night before, I had not realized how low the water level actually was. The waterline was about 10 m in from where it appeared to be on the ArcGIS map that was used to setup the course. Unfortunately, the WeatherBox  that housed the camera underneath AMOS struck a rock at full speed, and the 3-D printed extension piece that connected it to the main electronics enclosure snapped under the strain. The WeatherBox was quickly flooded, but I was following close behind in the kayak and was able to turn off AMOS's power switch a few seconds later. Clearly a more flexible extension piece will be required here. Perhaps something like a short rubber hose could work?

Despite this accident, the rest of the test was a success. Depth, temperature, and conductivity data were all recorded normally, AMOS navigated itself normally, and 9 video files were recorded, mostly of air bubbles and river vegetation, but at 1:10 of the video below, you can see 3 small fish swimming by for a few seconds:

The end of the above video is where the WeatherBox struck the rock and the camera module was flooded. Unlike the last camera flooding which occurred in salt water, this one was in fresh water, and after storing the camera module in a bag of desiccant for 12 hours, it was confirmed to be unharmed. 

Wednesday, September 1, 2021

AMOS Looks Underwater

 Last month's assumption that the o-ring seal on the WeatherBox required improvement proved to be correct. The printed part had tiny imperfections underneath the o-ring that allowed water to slowly leak through. A WeatherBox customer who was using it to observe a muskrat underwater came up with a brilliant solution: he found that using some marine goop (ex. in the channel holding the o-ring worked to fill in the small imperfections in the plastic and create a watertight seal. I've built a few WeatherBox enclosures since then with the channel filled with Marine Goop and they have all worked quite well.

To take pictures and video underwater with the 6 foot AMOS surfboard, an extension piece was required to get the WeatherBox below the waterline. The bow of the surfboard where the camera is located is pitched upward at about a 20 degree angle, so this requires a curved extension piece. The creation of this extension piece required a couple of weeks. At first, rendering the model in OpenSCAD took days for my laptop to finish, although later iterations of the model used some 2-D optimizations with extrusions that shortened the rendering time to about 24 hours. The first two model attempts also had small gaps on the side with the largest radius of curvature, which led to leaking. Eventually a working model of the curved extension piece was created:

The extension piece was fitted on AMOS and used to capture this underwater backyard pool video:

Apologies to viewers for the acting talent used in the above video. In Nature Robotics operates on a tight budget. 😉

Coming up next week is the final round of  the 2021 edition of the Ocean Startup Challenge ( In Nature Robotics is in the mix again this year, and will be pitching on Thursday, September 9.