Autonomous flying software: what do you use?

[Kyle76]

I've used Dronelink, but it's not compatible with the Air 2 (other than beta) currently. I have looked at Maven, and it looks like it might be a little simpler than Dronelink, and with a $14 price tag, I assume there are no additional charges for enterprise use? Dronelink gets quite pricey with its enterprise tiers. Any suggestions for simple waypoint missions with the A2?

 

[BigAl07]

What are your NEEDS in terms of Autonomous Flights?

 

[Kyle76]

There are some missions that I would like to save to use repeatedly over time, and some missions that it would be nice to have the ability to plan in advance and run autonomously on site, whether still or video. I hate that the A2 has no waypoint video capability, only hyperlapse, which seems extremely odd to me.

 

[BigAl07]

There are some missions that I would like to save to use repeatedly over time, and some missions that it would be nice to have the ability to plan in advance and run autonomously on site, whether still or video. I hate that the A2 has no waypoint video capability, only hyperlapse, which seems extremely odd to me.

It's possible those more complex options require more processing than the A2 is capable of handling. Keep in mind it's a budget friendly aircraft so they had to cut corners to get it to market in that price range and it's possible (likely) that processor and such are places they cut corners.

I have no experience with the MA2 but if it's compatible have you checked out Litchi?

 

[MapMaker53]

Litchi is compatible with: Mavic Mini 1, Mavic 2 (Zoom/Pro), Mavic (Air/Pro), Phantom 4 (Standard/Advanced/Pro/ProV2), Phantom 3 (Standard/4K/Advanced/Professional), Inspire 1 (X3/Z3/Pro/RAW), Inspire 2 and Spark.

There are some missions that I would like to save to use repeatedly over time, and some missions that it would be nice to have the ability to plan in advance and run autonomously on site, whether still or video. I hate that the A2 has no waypoint video capability, only hyperlapse, which seems extremely odd to me.

Litchi allows you to save waypoint missions so you can fly the same mission in the future. You can also use their Litchi Mission Hub website to pre-plan missions via your home computer and download them to your drone.

 

[adm_geomatics]

The topic of using the MA2 for mapping has been discussed on many forums and we have come to realize that the DJI SDK has not been given waypoint capabilities. This is a decision by DJI so at this point no autonomous flight software that uses waypoint functionality will be able to support the MA2. As an example it took them right at about a year to release the functionality to the Mavic Mini so the Mini 2 and the Air 2 are probably at least 6 months away, but I would be willing to bet that Litchi is the first one that adopts it.

On another point of using the MA2 for mapping people should be aware that it is only a 12 megapixel sensor and that the 48 megapixel effective spec that they state is essentially a stitched image that will not be useful for mapping.

 

[Bobstaff]

I am setting up my P4P on Litchi for waypoint flying, plus the Dronebase mapping licence only.
My Mavic 2 Enterprise Dual and Evo 2 both have mapping functions in their apps that work pretty well.

 

[Kyle76]

As an example it took them right at about a year to release the functionality to the Mavic Mini so the Mini 2 and the Air 2 are probably at least 6 months away, but I would be willing to bet that Litchi is the first one that adopts it.

On another point of using the MA2 for mapping people should be aware that it is only a 12 megapixel sensor and that the 48 megapixel effective spec that they state is essentially a stitched image that will not be useful for mapping.

I don’t think the Mini or Mini 2 have waypoint capability. The Air 2 does, but it is only for hyperlapse shots. It doesn’t seem like much of a jump from hyperlapse to full video in terms of processing power. As to mapping, I would think in most cases, barring obstacle avoidance, the Air 2 could just fly lower if greater resolution is required. For most ag mapping, it’s possible to avoid obstacles at 85 feet or so, at least in my area.

 

[JimD]

I don’t think the Mini or Mini 2 have waypoint capability. The Air 2 does, but it is only for hyperlapse shots. It doesn’t seem like much of a jump from hyperlapse to full video in terms of processing power. As to mapping, I would think in most cases, barring obstacle avoidance, the Air 2 could just fly lower if greater resolution is required. For most ag mapping, it’s possible to avoid obstacles at 85 feet or so, at least in my area.

The problem with just flying lower is that The field of view is much smaller and you have to fly much longer to cover the area.
When you’ve already got an 8 battery map do you really want to make it take longer?

Dronelink supports the MA2 It’s in beta because the DJI software is not fully ironed out yet.

 

[Ajkm]

Note the difference in meaning between the words "automated" and "autonomous". When referring to the latter, you are talking about a situation where the drone (or drone application) itself can make its own decisions about its operation when it encounters changes to the operational environment, without reference or input from the pilot.That means that an autonomous operation does not have 100% predictable outcomes. When talking about repeated identical flightpaths, one is referring to "automated" flight.

 

[JimD]

Note the difference in meaning between the words "automated" and "autonomous". When referring to the latter, you are talking about a situation where the drone (or drone application) itself can make its own decisions about its operation when it encounters changes to the operational environment, without reference or input from the pilot.That means that an autonomous operation does not have 100% predictable outcomes. When talking about repeated identical flightpaths, one is referring to "automated" flight.

Thanks for pointing that out it always kills me when they say autonomous. Because they’re not.

 

[Ajkm]

Thanks for pointing that out it always kills me when they say autonomous. Because they’re not.

I could have added the "goal-based" / "deterministic" systems distinction... where the deterministic system has a goal but the path to achieving that goal may vary and is rarely predictable, resulting in unique outcomes every time it's flown.

 

[adm_geomatics]

Note the difference in meaning between the words "automated" and "autonomous". When referring to the latter, you are talking about a situation where the drone (or drone application) itself can make its own decisions about its operation when it encounters changes to the operational environment, without reference or input from the pilot.That means that an autonomous operation does not have 100% predictable outcomes. When talking about repeated identical flightpaths, one is referring to "automated" flight.

Do you have any idea the engineering that is involved in a simple nadir mapping mission? The user inputs a desired altitude and area but flight calculations, terrain awareness, obstacle avoidance, the imu, altitude monitoring and camera configuration are all under the control of the software. Duty over desire is a big part of autonomy as well. We're not talking about AI in this sense. It is autonomous flight software in the case of mapping. Litchi waypoints on the other hand are much more basic and every characteristic is under the control of the user, but it is still using sensors to adjust its behaviour and to keep itself safe.

The Air 2 does, but it is only for hyperlapse shots.

Having "waypoint functionality" is a layman's term for what is actually involved in creating and running an autonomous flight plan, but how many "waypoints" can you use? Five? In mapping they have completely control of the drone the entire time including being able to record points in between waypoints of major intersection. In just a short 10 acre mission there can be at least hundreds of waypoints that are not only predetermined, but are also generated in accordance with the GPS signal and also when to take pictures, not just rolling a video. We won't even get into having a piece of software define true altitudes according to terrain awareness. The a completely different discussion of the SDK.

 

[JimD]

Do you have any idea the engineering that is involved in a simple nadir mapping mission? The user inputs a desired altitude and area but flight calculations, terrain awareness, obstacle avoidance, the imu, altitude monitoring and camera configuration are all under the control of the software. Duty over desire is a big part of autonomy as well. We're not talking about AI in this sense. It is autonomous flight software in the case of mapping. Litchi waypoints on the other hand are much more basic and every characteristic is under the control of the user, but it is still using sensors to adjust its behaviour and to keep itself safe.


Having "waypoint functionality" is a layman's term for what is actually involved in creating and running an autonomous flight plan, but how many "waypoints" can you use? Five? In mapping they have completely control of the drone the entire time including being able to record points in between waypoints of major intersection. In just a short 10 acre mission there can be at least hundreds of waypoints that are not only predetermined, but are also generated in accordance with the GPS signal and also when to take pictures, not just rolling a video. We won't even get into having a piece of software define true altitudes according to terrain awareness. The a completely different discussion of the SDK.

The technology in a drone was pretty much figured out decades ago. PID Loop was invented in 1911 and guidance system principles were figure it out in the 60s.
The Kalman filter which is the most complicated part Of a drones firmware was invented in 1961 and used in the space program

A cruise missile in the 80s could precisely hit a target thousands of miles away with no GPS.​

And all the technology is open source now, pretty much run of the mill. Nothing special at all.
As far as autonomy all it’s doing is taking its current GPS position and calculating the bearing of the desired GPS position and steering to there. That’s automation.

 

[R.Perry]

Interesting thought about automated or autonomous. To me when I do mapping, it is autonomous. There is no way I could manually fly the pattern and create the photographs with the correct overlap. To me that is an autonomous flight not an automated flight.
As for a hitting a target thousands of miles away without GPS, well that is true and you wouldn't want to rely on GPS to hit an enemy target. But reset assured the military had some very advanced ways of controlling ICBM back in the 70s and 80s.
GPS is a great tool, it has also made pilots lazy and too dependent on the current technology.

 

[JimD]

Back in the 80s I did some contracts for McDonald Douglas who are the prime contractors for the Tomahawk.
Basically what happened just before launch you download a mission plan that has the flight path and certain images that it looks for basically when you crossed over certain landmarks like from land to water it could pattern match the shoreline and figure out exactly where it was and re-calibrate the mechanical gyros and mechanical accelerometers.
It had a whopping large 10 MB drive that was high G-Shock proof. I wrote this one program called declassify that wrote patterns over the hard drive so if the enemy was coming down the tube you ran this program and it wiped everything.
The rep from McDonald Douglas would kid us He could call in a strike from St. Louis where McDonald Douglas was and it would hit my window and not his.
It’s amazing the things we did with such crude technology.

 

[adm_geomatics]

Back in the 80s I did some contracts for McDonald Douglas who are the prime contractors for the Tomahawk.
Basically what happened just before launch you download a mission plan that has the flight path and certain images that it looks for basically when you crossed over certain landmarks like from land to water it could pattern match the shoreline and figure out exactly where it was and re-calibrate the mechanical gyros and mechanical accelerometers.
It had a whopping large 10 MB drive that was high G-Shock proof. I wrote this one program called declassify that wrote patterns over the hard drive so if the enemy was coming down the tube you ran this program and it wiped everything.
The rep from McDonald Douglas would kid us He could call in a strike from St. Louis where McDonald Douglas was and it would hit my window and not his.
It’s amazing the things we did with such crude technology.

What does any of that have to do with modern drones and code that is completely different that even what was back in the 90's? So you believe that they aren't making any decisions based on anything like the altimeter, vision or OA sensors and that none of that has anything to do with the firmware and software? I'll just disagree that when a machine is making decisions like that based on no input from the user then that is autonomous. Despite how crude it may seem even a camera on Auto mode is making decisions.

 

[dougcjohn]

In my opinion sUAVs are automated, they carry out automated missions but still rely on human operators. However, in addition can be autonomous, by one definition source "UAV that can operate without any human intervention". In other words, autonomous drones take off, carry out missions, and land completely autonomously without human intervention... not all sUAV’s can be autonomous drones.

May involve higher interpretations for larger UAV Drones, intercontinental missions, higher matrix level of decisions based on additional program variables; but for this discussion, sUAV’s uploading a flight program and performing the flight without my interaction... other than observation and ready to abort if needed; it’s flying autonomously. Several software programs include the option if RF to radio is lost to 1) abort and run alternate program to RTH, 2) continue & complete flight mission and RTH at completion.

That’s autonomous, I’m not electing to override and control. I’ve created the mission, elected & input variables, saved mission (file), uploade mission and initiate the program; allowing the program and FW to interact and fly the mission. It is using GPS, although that is part of the programs parameters & FW SDK. It also includes actions to perform at designated points and between points using other sub-programs: camera settings, gimbal rotations, etc. Can also be programmed to run multiple operations; stop for period of time, run video then photograph. Some even fly a short distance to self finalize speed and camera settings prior to start of assigned mission. If thermal, can program a specific temp zone and perform additional programming, including notifying console of alerts or modifying flight mission. Included in autonomous can be automated portion too... example, if battery is low pause autonomous flight and run automated routine for battery swap... change battery and return to paused stage and continue autonomous programming.

The use of photography in mission... that’s close or now used... the Skydio S2 is getting new SW using it’s optical sensors in more than optical avoidance.

One area I’ve been disappointed is utilizing OA within some Software apps & missions. Many turn off or don’t use OA with autonomous flights... such as mapping programs; but that’s a limitation within SW design, not capability.

The Electronics on-board and processing power in newer platforms compared to even a few years ago is impressive. This allows advancement in additional automated and autonomous operations.

 

[adm_geomatics]

In my opinion sUAVs are automated, they carry out automated missions but still rely on human operators. However, in addition can be autonomous, by one definition source "UAV that can operate without any human intervention". In other words, autonomous drones take off, carry out missions, and land completely autonomously without human intervention... not all sUAV’s can be autonomous drones.

May involve higher interpretations for larger UAV Drones, intercontinental missions, higher matrix level of decisions based on additional program variables; but for this discussion, sUAV’s uploading a flight program and performing the flight without my interaction... other than observation and ready to abort if needed; it’s flying autonomously. Several software programs include the option if RF to radio is lost to 1) abort and run alternate program to RTH, 2) continue & complete flight mission and RTH at completion.

That’s autonomous, I’m not electing to override and control. I’ve created the mission, elected & input variables, saved mission (file), uploade mission and initiate the program; allowing the program and FW to interact and fly the mission. It is using GPS, although that is part of the programs parameters & FW SDK. It also includes actions to perform at designated points and between points using other sub-programs: camera settings, gimbal rotations, etc. Can also be programmed to run multiple operations; stop for period of time, run video then photograph. Some even fly a short distance to self finalize speed and camera settings prior to start of assigned mission. If thermal, can program a specific temp zone and perform additional programming, including notifying console of alerts or modifying flight mission. Included in autonomous can be automated portion too... example, if battery is low pause autonomous flight and run automated routine for battery swap... change battery and return to paused stage and continue autonomous programming.

The use of photography in mission... that’s close or now used... the Skydio S2 is getting new SW using it’s optical sensors in more than optical avoidance.

One area I’ve been disappointed is utilizing OA within some Software apps & missions. Many turn off or don’t use OA with autonomous flights... such as mapping programs; but that’s a limitation within SW design, not capability.

The Electronics on-board and processing power in newer platforms compared to even a few years ago is impressive. This allows advancement in additional automated and autonomous operations.

So if I were to buy into what you all are saying then my position is that there is no such thing as autonomy until we reach the level of AI that a machine can function as if it we a person and procreate or at least build themselves into a self-governing society. That's a bit daft guys. Just because autonomy means a certain level of sophistication in your mind does not mean that we aren't at a level of autonomous flight at this point. The Skydio R2 is 100% capable of scanning an obstacle filled environment without a preset pattern of it's own, but guess what - someone engineered the hardware, engineered the software and firmware and someone else told it what area to scan. A Tesla is told where to go. There is not a machine on this planet that doesn't exist because of a human and the same thing will remain unless the replicators from Star Trek or Stargate come about and even they were the construct of another alien race. I guess the thought of autonomy should be removed from machines. Ludicrous.

 

[dougcjohn]

So if I were to buy into what you all are saying then my position is that there is no such thing as autonomy until we reach the level of AI that a machine can function as if it we a person and procreate or at least build themselves into a self-governing society. That's a bit daft guys. Just because autonomy means a certain level of sophistication in your mind does not mean that were aren't at a level of autonomous flight at this point. The Skydio R2 is 100% capable of scanning an obstacle filled environment without a preset pattern of it's own, but guess what - someone engineered the hardware, engineered the software and firmware and someone else told it what area to scan. A Tesla is told where to go. There is not a machine on this planet that doesn't exist because of a human and the same thing will remain unless the replicators from Star Trek or Stargate come about and even they were the construct of another alien race. I guess the thought of autonomy should be removed from machines. Ludicrous.

You've quoted my post on that rant reply... I didn't suggest any of what you're indicating. Actually, my post was more in line with your previous comments. The apps used with sUAV is autonomous and the sUAV has automation procedures within FW.

In short... autonomously is operating without pilot intervention running programs and using onboard electronics & sensors.