Competition from skydio 2 is good for the entire consumer drone industry, activetrack is dji’s true achilles heel… | vimeo.com/382213949 | archive.Org/details/dji_m2p_activetrack_crash_dec312019 | I’ve been flying dji drones since the first mavic first came out back in late 2016. Since that time i’ve bought a spark, mini and also have the mavic 2 pro as well. Dji has pretty much dominated the consumer drone market for a couple years now, especially after the battle with gopro went decidedly in favor of dji. | Most people considered the mavic 1.0 state of the art at the time, and the mavic 2.0 an iteration on top of that. Before the skydio 2 came out, the dji activetrack 2.0 was pretty much considered the best there was on the market. | 2019 hadn’t been a good year for dji. I felt that they really didn’t come out with anything new or mind-blowing and in fact the mini was a step backwards. Sure the mini lets them market the mass lower market, but it demonstrated no progress in terms of technological advancement. The innovation train at dji seemed to have stopped or at least paused for a year. | While the skydio 2 as of yet is not really an all-around well rounded replacement of the mavic 2, it does have the competitive advantage in terms of true autonomy and ai tracking that puts dji’s current lineup to shame. | I personally don’t have a skydio yet but i’ve been watching a lot of youtube videos showcasing the competitive advantage of the skydio especially as compared to dji active track. Having had mavic for a while, i’ve never really tried its activetrack 2.0 features, but having some time on my hands on the last day of the decade Dec 31 2019) and living near a large high school with empty deserted parking lot i decided to give it a try to have it chase my car while i’m driving super slowly Less than 10 miles per hour). | Long story short, it ended in disaster with my mavic pro crashing and suffering much damage. This was my very first time using active track from dji, and it was also my very first crash out of hundreds of successful flights. The fateful flight in question was done in broad daylight, little to no wind, not in or even anywhere near any nfz zones or areas, and as well i was utilizing the “slow mode” of the activetrack 2.0 in which the drone was using all sensors including the side sensors. I didn’t subject it to tricky situations, no skinny tree branches or reflective mirrors or any gimmicks or gotchas. While it was actively tracking me in the slow mode it collided into an obstacle that it could have, and indeed should have been able to see in time to completely avoid. | What made the matter so much worse is the string of issues that lead up to the crash. It wasn’t just the fact that in this instance activetrack 2.0 failed me and caused the dji mavic to collide with an object that it should have avoided, it was the fact that this accident was a chain of issues that all lined up and each of the issues themselves was attributed to yet another failure on the part of dji itself. | You see, when i first set out to test activetrack 2.0 on my mavic pro 2 in the empty parking lot i was cognizant of the height issue and wanted to maintain a floor of at least 200 feet agl at all times, this was basically my backyard and my neighborhood and i knew that 200 feet would give me plenty of margin for error even if activetrack failed to see or avoid any taller obstacles. | However even though this was in broad daylight, with almost no winds that day, and i was the only car in the very large empty parking lot, when flying at around 200 feet in the air the mavic would lose activetrack within a few seconds after having acquired a lock and started tracking. In essence, activetrack was essentially nonfunctionality in terms of attempting to track a vehicle In this case my own car going very slowly at less than 10 mph at all times) at a height of just 200 feet. | So i had to lower the height to something much lower to have it actually be able to track my car. This was the first failure of dji activetrack not being reasonably robust enough and having these severe limitations that required the tracking to be commenced in such low altitudes that it put the aircraft in a significantly more dangerous flight envelope and also contributed to the lack of safety margins that ultimately combined with other flaws in the dji design that ultimately lead to fatal crash. ) | I lowered it to around 80 feet and it would track my car, but for the life of me i could not get it to stay in “behind” mode, i wanted it to track my car from behind and not from the front or from the side, but every time i started a track from “behind” it would inevitably and invariably move to the side the moment i turned my car in any direction. | There was simply no way i could force the mavic to keep tracking me from the behind, regardless of how slow i was going in my car. This was the second failure on the part of dji activetrack, as had it been able to follow simply instruction of tracking me from behind it could have compensated for the first flaw Requiring very low height in order to keep tracking lock) without putting safety in jeopardy, but since it decided to start moving to the side this put the mavic in a position where potentially it would be more prone to strike objects, given the dual prong facts that the side sensors were less able than the stereo front /back sensors, and also that when tracking a car on the street, usually the behind profile would be safer than the side profile in terms of flying/tracking at lower heights. This was basically an all-around bad combination of events that dji itself caused… | So after having spun a few times around the parking lot, i decided to drive a short block around my immediate neighborhood street. I drove very slowly, slower than 10 miles per hour, and as i turned the corner i could see that since the mavic choose on its own to shift to the right side tracking profile it would come within close range to a taller tower-like structure or obstacle that we were approaching. | Immediately i grabbed hold of the joystick and made inputs for the mavic to increase in height. I was still getting a perfect real-time picture feed and perfect downlink from the mavic cameras to my dji go app on my ipad tablet device connected to the mavic remote controller, and at no point did i have any loss of connection or loss of video feed, however i observed that despite my full deflection on the height input telling the mavic to climb as quickly as it could to get as much altitude as it can, the mavic stayed in its activetrack profile and refused my command inputs to increase in altitude. | Out of all my years of flying the mavic i’ve never encountered any situation in which it would refuse to listen to my input to command it to increase in altitude. The only other scenario in which this happened was when it was in a nfz with a max altitude/height restriction, but this being in my own backyard i knew for sure 100% that there were no nfz’s around nor any altitude restrictions, and in any case, my mavic was only about 100 feet at the time, well under the 400 feet agl as dictated by the faa. | The only thing i can surmise is that the activetrack 2.0 had a glitch or a bug that caused in this instance the mavic to refuse all height input commands. This reminds me of the boeing 737max mcas issue in which the pilots try to save the aircraft but the flawwed ai thinks it is smarter than the pilot and then locks the pilot out, causing a fatal crashes. | Soonafter, before i even had a chance to manually cancel the activetrack and take full manual control, i heard a loud band and a thud noise and i knew the mavic had collided with something while flying in the “slow” Safer) mode of activetrack, in broad daylight, with no wind, nowhere near a nfz zone, and with all sensors including side sensors actively operating. Not only did the mavic refuse my command input to increase height, not only did its obstacle avoidance completely fail to avoid the object, but it hit the object while refusing to allow me to make manual corrections that could have saved the day. | My next course of action after i got home with the damaged drone was to figure out what the heck happened. Part of that was the recovery of the flight logs, but as it turns out, the flight logs were mysteriously missing and apparently deleted by the dji go app itself. It wasn’t like i flew a whole bunch of missions after this and the flight logs somehow overwrote itself, the fact of the matter was i didn’t fly any missions after the crash and didn’t even touch the ipad until it was time for me to attempt to retrieve the logs to figure out what the heck had happened. So the only possible scenario in which the flight recorder logs disappeared was that dji simply didn’t log it in the first place, or somehow the dji app locally or dji themselves remotely deleted it somehow after the fact. This was more mysterious given the fact that i had logs from the 24th, and so the other older missions logs were still intact but not the mission that was most critical for analysis! | I also learned that even though the actual raw .Dat logs were on the ipad and could be retrieved that starting with the mavic 2 dji decided to encrypt those logs without giving anyone the key. So the .Dat logs were for all intents and purposes useless to anyone but dji themselves, how transparent of them! | One would think in this day and age in which the us government is deciding whether or not to ban dji Much like how they banned huawei etc) that dji would be more transparent and would choose to not encrypt the logs! for one thing, we all know dji is coming up with app to allow anyone to track dji drones anywhere, and the dji drones is emitting telemetry to the entire world and public at large, so from end user Customer/drone pilot) standpoint dji seems to think that level of sharing privacy is okay but then dji goes behind the backs of its customers and encrypts their own flight log data and doesn’t give the key?! | It just seems backwards… Dji actively disregards the privacy of their own customers but then go out of their way to hold their own customers’ data hostage from themselves! even critical diagnostic data! this makes no sense whatsoever and doesn’t even pass the smell or sniff test. | This is not what a customer service oriented company should treat its paying customers. This is not a good trend for the industry! trade war or no trade war, entity list or no entity list, dji should not be eroding the privacy of their paying customers whilst at the very same time locking their customers down and blocking them from access to their very own data! there is no excuse for this double standard and hypocrisy! | I’m not anti-dji and in fact in the past when i criticized other drones like the uk zano drone i was publicly accused of being an employee of dji : | For the past couple years, i’ve been a huge supporter of dji and their products, but there comes a point in which one has to admit that dji is starting to fall behind in terms of innovation, and without a credible contender to effectively compete with dji, then i do believe we are already seeing stagnation in the industry and this is to the detriment of the end users and the consumers and the growth and viability of the industry itself long term. | It is not surprising then that skydio 2 has targeted its strengths to be exactly where dji is lacking the most, in terms of true ai autonomy and real active tracking… While the dji mavic 2 is still a very good product overall, and overall well-polished, its biggest lack would definitely be its glaring weakness when it comes to things like autonomy, active track and other forms of ai modes, or lack thereof. | I’ve flown the mavic for years, hundreds of flights, never a hitch. That is a testament to the quality of the dji mavic product. But the very first time i attempt an active track flight, dji flies the mavic into an object and crashes. Sure this is just one instance and not representative nor statistically significant, but it is significant to me subjectively and it shows me that when it comes to things like activetrack, dji products are not to be trusted and still light-years away from any sort of truly smart autonomy or trustworthy tracking capabilities. | It is my hope that the third version of the mavic Mavic 3.0) will greatly improve in this regard, but as things stand, the mavic 2.0 activetrack 2.0 is in my opinion less than useless and it almost cost me a loss of nearly $1600 usd. This is at best a total loss of confidence in dji’s so-called activetrack tech and at worst almost criminal level of false advertising. Since i was using the “slow mode” which was pitched as being full sensor ultra-safe, and this was flown under the most ideal of circumstances, broad daylight, super slow, no wind, fully up to date firmware, etc) | To sum it up, we as consumers should welcome competition in the open market from all players and all angles, if skydio 2 can bring the best out of dji and finally force them to make mavic 3 with actually usable active track and real autonomy then all the better, but if dji fails to improve then i can foresee the day in which a skydio 3 could be a full on replacement for any dji mavic or other dji drones in the future.
Country: United States
City: Los Angeles
Address: 6424 Santa Monica Blvd