Emergency landing with no control

Sometimes, you are lucky. And sometimes very very lucky. Today I tested software (Litchi) and hardware (Mavic 3). As a result, I had a horrible experience lasting 20 minutes, but I learned a lot. And it ended well. The purpose of testing is learning. And you learn best from mistakes, because you can do better afterwards. A flawlessly executed test is useful because it confirms that a test run with a specific setting went fine. But several near disasters is even better, because it creates more learning experiences. So, what did I do (wrong)? I started with testing Litchi. Litchi is an app that provides functionality that the DJI app lacks. The Litchi software enables you to let the drone have an autonomous flight. The drone can fly itself without interaction via the controller by executing instructions that are uploaded into the drone before take-off. These instructions are a program that can be created offline in the Litchi website, via a intuitive interface. Basically, you point and click on a Google Map, and create waypoints. At each waypoint, you can program what the drone does: head for a next waypoint, or stop, of take a photo for example. There are a lot of parameters, but the basics are simple. During the test, I let the drone fly over the outlines of a former fort and a museum. The total flight took about 10 minutes, distance up one kilometer. I programmed the drone to fly straight above specific landmarks, in that way I tested the navigation precision. The drone flew exactly as planned, and the horizontal difference between planned location on the map, and actual position shown on drone images, was about 1 or 2 meters. In the picture above, the deviation size can be estimated by comparing it the the broom on the right: less than 1 meter. Litchi is precise. The only things that did go different from planned, was that I programmed the drone to take photos, which didn't work, probably because I also programmed the drone to make video footage. Probably, I need to stop the video in an instruction in a waypoint in order to have the drone make photos. Also, The drone does not keep focussed on viewpoints, unless explicitly programmed to do so. For the purpose that I need Litchi for, it works perfect. The second test involved flying with the Mavic 3. It was the first time I flew the Mavic at a longer distance. I started by testing the drone at 5 meter height, it worked fine. I tried to make photos at 10 meter, and I noticed the drone drifted. I immediately understood that the drone used visual positioning, and when flying above moving water, this results in an unstable flight. When I moved the drone above shore, the drifting stopped. Check: works as designed. After that, I flew the drone up to its range limit. The drone complained about bad connection, as expected. And when disconnected, it started a return to the home point, as expected. I stood up, got a better connection, and pushed the drone further out. Then, my iPad stopped working. I quickly understood that the cold had made it stop working, I had previous experiences with iPhones in very cold weather. I clicked the return-to-home button on the controller and started heating my iPad. I also tried to connect my iPhone12, but somehow this phone does not connect to the Mavic 3 controller. So far, no stress. But then I slowly started to realise that the drone did not return to sight within the expected few minutes. And I remembered that I did not see the "homepoint has been recorded" message after starting the drone. That implied that the drone would not return to the point where it started, but to another homepoint, probably somewhere between the starting point and the point far out. Stress. I kept warming the iPad and trying to start it up by repeatedly pressing the side button. No response. I biked to the furthest point where the drone went. No hovering drone. I biked back to the start position. No landed drone. Further, about 15 minutes after loosing connection, the controller started beeping as if the drone had a low battery. And some moments later, it beeped unstoppable like it was making an emergency landing. Only after that, I finally realised that an iPad is not started by simply pressing a button, but by keep pressing the button until the Apple logo is visible. When that happened, I was able to restart the DJI Fly app, but it did not show anything, no connected drone. Again, I biked back to the area where the drone could be. And to my surprise, the DJI fly app showed connection. After selecting "find my drone" it showed the location of the drone on a rudimentary map, and it showed the image from the drone: it had landed in an area with high grass. It also showed that the drone battery had 2 percent left. I used the button to let the drone make noise, and based on the sound, I could locate it on an area with grass, ten meters from trees, three meters from water. It had landed safely after draining the battery almost completely and making an uncontrolled emergency landing. I was happy. My faults (learning points) 1. I did not wait for the "homepoint has been recorded" before flying out, 2. I did not anticipate my iPad to get too cold and stop working, 3. I did not have a backup device at hand, 4. I forgot how to restart my iPad. Lessons learned: don't make four mistakes in a row