Welcome, Commercial Drone Pilots!
Join our growing community today!
Sign up

Good review of Skydio 2 tracking by DC Rainmaker (cyclist reviewer)

dougcjohn

Well-Known Member
Joined
Jan 17, 2018
Messages
519
Reaction score
195
Age
66
DC Rainmaker is a popular reviewer & channel in the cyclist world.
He’s known for through, in-depth... honest reviews of pretty much any thing that can be used by Cyclist or Running sports.

He’s done a nice review of the S2 in various density of real life active tracking.
Thought it‘s been the best one so-far showing actual usage, not propped looking.

 
  • Like
Reactions: TreeLineView
For a comparison... of DJI’s best offering “Active Track 2”, by the same Reviewer, DC Rainmaker... you can see how S2 has the superior tracking system

 
I'd agree, it's probably not the cinematic smooth flow platform, and I don't think the I2 X7, or various Octal / Hex are threatened. But for just the active track component... not quite so thick trees (all videos, more in companion thread on Skydio forum), it does very well.

I also believe you can easily set altitude and position... again, in a little more open space looks to work great.

I still feeling the S2's main audience is going to be solo riders/runners/ski/etc are going to capture themselves. It's Beacon will be primary, the RC & App... and video screen will be secondary.

For a pure sports sole autonomous, it's tracking easily outperforms the M2P.

The M2P's tracking method that allows PIC input and control gives a smooth near 2-person operation.
 
Of you’ve followed various product threads at RCG you might have run across a guy called Vegasrobbie. He either owns or runs a custom off road vehicle outfit somewhere in Nevada and he makes extensive use of follow, active track functions with various brands to showcase his work and products. You can review his stuff on YouTube.

Where follow modes and obstacle avoidance is concerned I doubt very seriously if anyone has put all the models through more in field testing than he has. When the S-1 came out he ran it hard and put it away wet with his assessment having the S-1 with the best OA than any other by a significant amount.

If the S-2 does even better there probably isn’t any competition worth mentioning where OA is concerned.
 
Of you’ve followed various product threads at RCG you might have run across a guy called Vegasrobbie. He either owns or runs a custom off road vehicle outfit somewhere in Nevada and he makes extensive use of follow, active track functions with various brands to showcase his work and products. You can review his stuff on YouTube.

Where follow modes and obstacle avoidance is concerned I doubt very seriously if anyone has put all the models through more in field testing than he has. When the S-1 came out he ran it hard and put it away wet with his assessment having the S-1 with the best OA than any other by a significant amount.

If the S-2 does even better there probably isn’t any competition worth mentioning where OA is concerned.
I've seen a few of his on the Skydio sister forum. Along with his, and several others making offroad trail promo videos. He also used the Typhoon H too... great collection. Although most were vehicle sized paths and pretty much open range territory.

I found the tight bike trails & turns displayed the "tracking" strengths & weaknesses. Even with the beacon, the tighter dense trails still presented a hard time navigating the narrow trail with small twigs. Although technologicaly, not cinematic... It's impressive how it will often "find" a route through with a few "trial & error" attempts; although not at a pace to be useful to record the activities.

The section I found that displayed it's technology the best was the bike trail along the river with a row of trees on both sides. That was smooth, and able to change trail sides while riding the bike at a fast pace over 20mph... meaning not much attention to the S2 and a pretty fast pace in tree lines. Both the trees and bridge on trail displayed where the S2's OA excels. Solo person can create impressive video, similar to current 2-person creations.

The area where it clearly beats the M2P is due to the beacon on person... similar to VegasRobbie's use of the H's Magicwand. Passing behind brush or buildings, M2P predicts your unseen path, where the S2 tracks via beacon... that works better. Add in the S2 optical avoidance and processing matrix, the S2 does very well... but still not fast enough for tight trails. It was mentioned in one of DC Rainmakers video his conversation with Skydio and how the optical, as with 1st Skydio version, uses only 1 meter distance which is probably what effects the thicker trails... not looking far enough forward to reroute smoother.

Overall, it looks to be a major advancement in OA. It's not perfect... in some situations it'll challenge or beat human control, and in others won't approach the smoothness of manual control or advanced decisions on routing.

Next generation OA technology, might be similar to method of Human that visually inspects the intended flight path, essentially a pre-flight walking & inspecting route. If they incorporate a pre-flight memory, and allow the pilot to pre-fly the route so it can record technical challenging areas, an autonomous flight with adv OA would become very impressive.
 
Last edited:
Every system needs time to process obstacle information so I believe that’s where flight speed is reduced to provide it. Full scale military leaves high speed, low level obstacle avoidance to the human brain and human control input but even that encounters moments of over load. Too much info and too little processing time and we end up with crashes such as the F-18 terrain impact in Star Wars Canyon a few months ago.

Ultimately I think there’s a finite limit to how fast drones will be able to self navigate flight paths around obstacles. We may want higher speeds but such may not be possible.
 
I'd fully agree, we're probably not going to replace the human element with AI & autonomous flight programming... at least we hope so! Hmmm... didn't we say that for vehicles too about 5 yrs ago? Road surfaces, accident prevention, route planning, etc all programmed to a functional level & safety. Now planning semi-trucks going cross country driverless and cars driving 99.5% of the journey. As AI & processors cost get lower... We may be impressed. Personally, as an older model; I'm not having a car drive me and it'll be a day never approached where I fly in a pilotless craft.

Although backup to the 50's-60's, and what we thought would never happen in AI is unfathomable to us previously in reality. In all subjects including aviation... many aircraft wouldn't be flying now if it wasn't for electronics and the power of processing & computations between various circuits. Which actually, in comparison to modern hardware isn't very powerful processing requirements...and space & aviation still use lower scale computational hardware compared to modern computers, although with 3X to 6X failsafe redundant systems. Those fighter's armament systems are worthless without the controlling computers... a pilot wouldn't be hitting the mark with these persuasive intelligent ordinances without the processing power of AI and logic matrices. Navel vessels, and some defensive systems on military aircraft now have gun & rocket targeting systems that can clearly out perform human mental capacity to target & hit multiple incoming aerial salvos, making best tactical decisions in nano-seconds. Agreed this is a different dimensional element, but it's processing & computational powers thought impossible a few decades ago.

In addition, regarding SAR & improved Object detection & avoidance; not far from letting a Thermal & various detection sensors taking control in SAR missions. Capturing something worth examining as a probably target and ruling out time wasting targets based on the scaning details of multiple sensors while quickly covering the surface. That may approach sooner than realized the ability to exceed the human rate & capacity to assimilate the data as speed is increased to meet newly developed sensors.

Back to our dirt trails, trees and obstacles... increasing an small autonomous drone from 2-3mph navigating small obstacles to 7-10mph probably won't be overly challenging over the next 2-3 generations of smart drones. Almost there now on open terrain trails or more spacious obstacles, I can easily see the S2 with improved FW or version 3 S3 (or competitors) being able to track at their max craft's speeds of 30-60mph, depending on design and accuracy of pocket transmitter.

Long way from perfect... but approaching.
 

Members online

Forum statistics

Threads
4,291
Messages
37,658
Members
5,989
Latest member
AlanzFPV