Just now beginning to tune both gimbals (and tune the assoicated dampening systems).
Here's some early test video from the first two gimbal systems:
Thursday, June 20, 2013
Packing for the Bush
I had to sort through how to pack all the UAS components in a way that facilitated relatively easy transport, but also allowed for very quick forward deplopyment. The goal was a system that could be airborne within 5mins of arrival at any location. Step one was a smaller, more portable ground station. Here's V2.0 of the GCS:
Now, smaller, easier to carry, and designed in a water proof Pelican case.
Note the checklist to the right, now used before and during every flight, which was developed over several months of flight testing (i.e. anytime something nearly, or actually resulted in an incident or accident, the checklist was updated :) ).
Packing the UAV required some trade-offs. If the landing skids are disasembled, the whole UAV can fit in the single fairly small silver case to the right of this photo, but would take 15 mins to deploy (and introduce risk as it might be easy to miss a bolt while assembling in the field. I opted to use two cases for transporting the UAV:
The top section of the UAV is packed in the OEM case. The landing skids are packed, completely assembled with one of three mission gimbal/camera combinations in the left case.
The UAV can be assembled, and completely pre-flighted, along with the GCS in under 5 mins.
Batteries and associated peripherals got there own case for safety reasons. This case holds enough batteries for upwards of 3 hours of flight time!
The handheld controllers also have a dedicated case (along with room for spare cameras, lenses, etc)
I had to sort through how to pack all the UAS components in a way that facilitated relatively easy transport, but also allowed for very quick forward deplopyment. The goal was a system that could be airborne within 5mins of arrival at any location. Step one was a smaller, more portable ground station. Here's V2.0 of the GCS:
Now, smaller, easier to carry, and designed in a water proof Pelican case.
Note the checklist to the right, now used before and during every flight, which was developed over several months of flight testing (i.e. anytime something nearly, or actually resulted in an incident or accident, the checklist was updated :) ).
Packing the UAV required some trade-offs. If the landing skids are disasembled, the whole UAV can fit in the single fairly small silver case to the right of this photo, but would take 15 mins to deploy (and introduce risk as it might be easy to miss a bolt while assembling in the field. I opted to use two cases for transporting the UAV:
The case in the right of this photo is an OEM design that can hold the complete UAV if the landing skids are disassembled
Wednesday, June 19, 2013
Primary Mission Camera
The main mission cameras we're using are Sony NEX's (NEX5s and NEX7s). I needed a very high quality gimbal that not only supported the NEX line of cameras, but had the capacity to handle bigger DSLRs and camcorders if need be. It was surprisingly hard to find an off-the-shelf BLDC-based solution, so I ended up have one custom fabricated.
Note the integrated avionics package mounted on the rear of the gimbal. It includes necessities like onboard HDMI to composite video conversion, and two separate DC busses to power the various components of the camera, gimbal and downlink systems.
The main mission cameras we're using are Sony NEX's (NEX5s and NEX7s). I needed a very high quality gimbal that not only supported the NEX line of cameras, but had the capacity to handle bigger DSLRs and camcorders if need be. It was surprisingly hard to find an off-the-shelf BLDC-based solution, so I ended up have one custom fabricated.
Note the integrated avionics package mounted on the rear of the gimbal. It includes necessities like onboard HDMI to composite video conversion, and two separate DC busses to power the various components of the camera, gimbal and downlink systems.
Cameras, Gimbals, Mission Avionics
After several weeks of flight testing and tuning, I finally had a fully mission-capable end-to-end system, which included the UAV and both the Handheld Controller (i.e. RC controller integrated with handheld video) and a full blown Ground Control System.
It was time to start integrating the mission gimbals and cameras. The first of three camera/gimbal combinations was the GoPro with a direct drive brushless DC motor based Gimbal (using a SimpleBGC based gimbal controller).
After several weeks of flight testing and tuning, I finally had a fully mission-capable end-to-end system, which included the UAV and both the Handheld Controller (i.e. RC controller integrated with handheld video) and a full blown Ground Control System.
It was time to start integrating the mission gimbals and cameras. The first of three camera/gimbal combinations was the GoPro with a direct drive brushless DC motor based Gimbal (using a SimpleBGC based gimbal controller).
To make it easy to swap camera/gimbal combinations in the field, I decided to dedicate a landing skip set (DJI calls them BiPods) to each camera/gimbal combination because BiPods can be swapped off the S800 with 90 seconds, with no tools required.
Also included are all the DC buses and video downlink subsystems (which vary from camera/gimbal combination to camera/gimbal combination).
The GoPro mount is the simpliest, although it still has two indepentend DC busses and a 5.8Ghz xmitter.
Also included are all the DC buses and video downlink subsystems (which vary from camera/gimbal combination to camera/gimbal combination).
The GoPro mount is the simpliest, although it still has two indepentend DC busses and a 5.8Ghz xmitter.
Heres the 1st BiPod, configured with the GoPro mount:
Developing the Production 'Copter
After proving out the end-to-end flight control SW, systems and procedures, it was time to begin integration testing with the mission platform, a modified DJI S800.
I had an S800 built by a local shop using an off-the shelf Flight Controller (a DJI Wookong) and did a bunch of baseline testing with it (endurance, climb rate vs battery charge, stability-vs-CG, etc, etc).
Next, I torn down the S800, rebuilding it with an Arducopter-based flight control system, and developing a custom mounting system that would ultilmately support the range of camera/gimbal/avionics packages I was planning to develop. First step was getting an S800 flying with an APM. If you look carefully, you'll see the rudimentary camera gimbal and GoPro developed for testing both the platform, as will as the mounting system.
After proving out the end-to-end flight control SW, systems and procedures, it was time to begin integration testing with the mission platform, a modified DJI S800.
I had an S800 built by a local shop using an off-the shelf Flight Controller (a DJI Wookong) and did a bunch of baseline testing with it (endurance, climb rate vs battery charge, stability-vs-CG, etc, etc).
Next, I torn down the S800, rebuilding it with an Arducopter-based flight control system, and developing a custom mounting system that would ultilmately support the range of camera/gimbal/avionics packages I was planning to develop. First step was getting an S800 flying with an APM. If you look carefully, you'll see the rudimentary camera gimbal and GoPro developed for testing both the platform, as will as the mounting system.
Battery Charging
It didn't take many sortes to realize battery logistics required some planning and development.
I decided to introduce field re-charging early in R&D process so I could develop processes and procedures, and ultimately a platform for managing batteries in forward deployments.
Even though my R&D copter used relatively small 3 and 4 cell batteries, I had a system built that was capable of rapid charging the 8000MaH monster batteries I suspected I would ultimately be using. I did this because I also needed more sophisticated computer aided battery charging tech early in the R&D process (for accurate performance and endurance characterization), and figured I might as well design for the ultimate field deployment.
Here's the 2000W portable dual charging system I had built. Not only was it helpful for my early testing (produces tons of useful data), but this portable system has the capacity to quickly charge multiple of the 6 cell 8000MaH monster batteries I eventually ended up using with the mission 'copter:
I combined this with an portable generator to enable multiple sortes while out on the test range (which has no access to power)
It didn't take many sortes to realize battery logistics required some planning and development.
I decided to introduce field re-charging early in R&D process so I could develop processes and procedures, and ultimately a platform for managing batteries in forward deployments.
Even though my R&D copter used relatively small 3 and 4 cell batteries, I had a system built that was capable of rapid charging the 8000MaH monster batteries I suspected I would ultimately be using. I did this because I also needed more sophisticated computer aided battery charging tech early in the R&D process (for accurate performance and endurance characterization), and figured I might as well design for the ultimate field deployment.
Here's the 2000W portable dual charging system I had built. Not only was it helpful for my early testing (produces tons of useful data), but this portable system has the capacity to quickly charge multiple of the 6 cell 8000MaH monster batteries I eventually ended up using with the mission 'copter:
I combined this with an portable generator to enable multiple sortes while out on the test range (which has no access to power)
Developing Ground Control Infrastructure
Initial tests of the FW550+APM+Video downlink were run to understand both manual control, and rudimentary auto flight. Here's the handheld controller I used in these early tests, along with an attached video down link system I developed:
After early flight tests, it was time for more sophisticated automated flight tests, including guided (aka semi-autonomous) and fully autonomous modes.
I developed v1.0 of my portable Ground Control Station (GCS) to facilitate these tests. With integration of a bunch of peripherals via internal power buses, video and USB, this GCS enabled in-flight control of all aspects of UAV operation. It provided real-time mapping of vehicle location (using Google Earth Maps), access to real-time telemetry data, video feed, flight plan programing, vehicle mode control, etc, etc.
Over the course of several weeks, I was able to gradually expand the mission envelope to include both semi autonomous, and fully autonomous (i.e. autonmous execution of multiple waypoint flight plans) operation of the R&D copter.
Some early video from field tests of the end-to-end platform....
Initial tests of the FW550+APM+Video downlink were run to understand both manual control, and rudimentary auto flight. Here's the handheld controller I used in these early tests, along with an attached video down link system I developed:
After early flight tests, it was time for more sophisticated automated flight tests, including guided (aka semi-autonomous) and fully autonomous modes.
I developed v1.0 of my portable Ground Control Station (GCS) to facilitate these tests. With integration of a bunch of peripherals via internal power buses, video and USB, this GCS enabled in-flight control of all aspects of UAV operation. It provided real-time mapping of vehicle location (using Google Earth Maps), access to real-time telemetry data, video feed, flight plan programing, vehicle mode control, etc, etc.
Over the course of several weeks, I was able to gradually expand the mission envelope to include both semi autonomous, and fully autonomous (i.e. autonmous execution of multiple waypoint flight plans) operation of the R&D copter.
Some early video from field tests of the end-to-end platform....
Developing a flying R&D Platform
(and an education in the dynamics of multi-rotor flight)
The FW550 was fairly heavily modified, starting with a fairly flexible undercarriage system that had lots of mount points, and the addition of a rudimentary servo-driven gimbal for a GoPro:
Also included were 2 DC buses (in addition to both the ship's main battery bus, and the Flight Controller's avionics bus. I added both 12.5 VDC and 5VDC DC-to-DC converters, which enabled power for a long range video downlink (which need ~12V) and both the GoPro and Gimbal servos, both of which need fairly high current 5VDC.
By separating out these buses, I ensured the flight critical avionics bus was never affected by any payload-related issues, and also isolated the mission payloads from the ships bus (so I could experiment with different cell-count batteries on the ship's bus without having to reconfigure video downlinks, etc).
The key change made to my R&D platform was the addition of an Arducopter Flight Controller and associated telemetry.
My intention at this point was to begin exploring the various flight modes and expanding the mission envelope, but I soon learned I wasn't quite ready yet.
Despite the fact that the FW550 seemed to fly reasonably well, I soon learned all of the modifications I made created a platform that weighed much more than a stock FW550 could _safely_ lift, especially with the associated change in CG of hanging so far underneath the "lift disc" (one of the configurations my custom landing gear enabled were dual, under slung batteries, which create a wonderful endurance profile, but moved the CG well below the plane of the props).
Initially, it appeared I had no problem flying at this weight/CG, and in fact I flew quite a few early tests before gaining a huge lesson in the dynamics of multi-rotor flight.
I learned this lesson "Empirically" via two "incidents." First, I had a hard landing due to "settling with power." After analyzing what happen (perhaps a bit too superficially) via the telemetry logs, I determined that I could avoid similar incidents by carefully limited descent rates.
A key operational change was the institution of a policy where by I never descended under manual control above 20 meters because its too hard to visually judge descent rate that far out. After this incident, above 20 meters, I always used altitude "hold" mode with a descent commanded via either the handheld controller or the GCS (a policy that continues to this day). This eliminated the simple case of settling with power, but actually masked the larger issue I had with this configuration.
What I didn't fully understand, until after the next much more significant incident, (i.e. a fairly spectacular crash) was that I had created a platform with insufficient excess lift margin for the vehicle to maintain controlled flight during a variety of flight conditions.
The second incident was precipitated by a fairly slow descent through downdraft during a wind gust. While the slow descent rate kept the props out of the bulk of the down wash plum, there was still enough turbulence to cause some pitch and roll moment.
During one of these excursions, a wind gust compounded the rolling moment, creating a moment beyond which the platform had enough excess lift to compensate with. As a result, the vehicle reached an pitch/roll angle of 45 degrees, at which point the flight controller shut off all the motors. The rolling moment was strong enough to cause the roll/pitch to continue past 45 degrees, eliminating the possibility of motor restart.
The 'copter flipped on its back and fell to the ground.
The accident cycle happened so quickly it was hard to really see what had happened in real time.
Examination of the video and telemetry logs, however, revealed what had happened in gory detail.
After this second accident (the pictures of which I will spare everyone, not much was left to look at) I spent meaningful time examining the onboard video and logs, as well as video from my ground camera (and indeed I had lots of time to reflect because, unlike the hard landing, this accident left me with nothing to fly for a while).
I wasn't happy about loosing the copter, but gained tons of insight into how to properly engineer the battery-motor-prop combination to produce sufficient lift margins at a given weight.
After calculating the lift I needed, and developing and motor-prop combination that produced enough lift, I revisited my R&D vehicle decision.
I realized I either needed something bigger than my stock FW550, because I needed to use 11in props to get enough lift margin (the arms on a 550 are not long enough for anything more than 10in props). I needed to either find a different R&D platform, or make some significant mods to the 550.
For a bunch of reasons, I decided to stay with an (albeit heavily modified) FW550.
I added arm extensions, and put on larger motors and props, which created the requisite lift capacity to ensure I could maintain vehicle control throughout the expected flight conditions.
Here's the reborn, heavily modified FW550 that ultimately became my R&D workhorse.
After spending several days (somewhat nervously) probing the flight envelope, I was able to verify I finally had a flying dev platform capably of exploring the expected flight envelope carrying the payloads I needed to carry.
(and an education in the dynamics of multi-rotor flight)
The FW550 was fairly heavily modified, starting with a fairly flexible undercarriage system that had lots of mount points, and the addition of a rudimentary servo-driven gimbal for a GoPro:
Also included were 2 DC buses (in addition to both the ship's main battery bus, and the Flight Controller's avionics bus. I added both 12.5 VDC and 5VDC DC-to-DC converters, which enabled power for a long range video downlink (which need ~12V) and both the GoPro and Gimbal servos, both of which need fairly high current 5VDC.
By separating out these buses, I ensured the flight critical avionics bus was never affected by any payload-related issues, and also isolated the mission payloads from the ships bus (so I could experiment with different cell-count batteries on the ship's bus without having to reconfigure video downlinks, etc).
The key change made to my R&D platform was the addition of an Arducopter Flight Controller and associated telemetry.
My intention at this point was to begin exploring the various flight modes and expanding the mission envelope, but I soon learned I wasn't quite ready yet.
Despite the fact that the FW550 seemed to fly reasonably well, I soon learned all of the modifications I made created a platform that weighed much more than a stock FW550 could _safely_ lift, especially with the associated change in CG of hanging so far underneath the "lift disc" (one of the configurations my custom landing gear enabled were dual, under slung batteries, which create a wonderful endurance profile, but moved the CG well below the plane of the props).
Initially, it appeared I had no problem flying at this weight/CG, and in fact I flew quite a few early tests before gaining a huge lesson in the dynamics of multi-rotor flight.
I learned this lesson "Empirically" via two "incidents." First, I had a hard landing due to "settling with power." After analyzing what happen (perhaps a bit too superficially) via the telemetry logs, I determined that I could avoid similar incidents by carefully limited descent rates.
A key operational change was the institution of a policy where by I never descended under manual control above 20 meters because its too hard to visually judge descent rate that far out. After this incident, above 20 meters, I always used altitude "hold" mode with a descent commanded via either the handheld controller or the GCS (a policy that continues to this day). This eliminated the simple case of settling with power, but actually masked the larger issue I had with this configuration.
What I didn't fully understand, until after the next much more significant incident, (i.e. a fairly spectacular crash) was that I had created a platform with insufficient excess lift margin for the vehicle to maintain controlled flight during a variety of flight conditions.
The second incident was precipitated by a fairly slow descent through downdraft during a wind gust. While the slow descent rate kept the props out of the bulk of the down wash plum, there was still enough turbulence to cause some pitch and roll moment.
During one of these excursions, a wind gust compounded the rolling moment, creating a moment beyond which the platform had enough excess lift to compensate with. As a result, the vehicle reached an pitch/roll angle of 45 degrees, at which point the flight controller shut off all the motors. The rolling moment was strong enough to cause the roll/pitch to continue past 45 degrees, eliminating the possibility of motor restart.
The 'copter flipped on its back and fell to the ground.
The accident cycle happened so quickly it was hard to really see what had happened in real time.
Examination of the video and telemetry logs, however, revealed what had happened in gory detail.
After this second accident (the pictures of which I will spare everyone, not much was left to look at) I spent meaningful time examining the onboard video and logs, as well as video from my ground camera (and indeed I had lots of time to reflect because, unlike the hard landing, this accident left me with nothing to fly for a while).
I wasn't happy about loosing the copter, but gained tons of insight into how to properly engineer the battery-motor-prop combination to produce sufficient lift margins at a given weight.
After calculating the lift I needed, and developing and motor-prop combination that produced enough lift, I revisited my R&D vehicle decision.
I realized I either needed something bigger than my stock FW550, because I needed to use 11in props to get enough lift margin (the arms on a 550 are not long enough for anything more than 10in props). I needed to either find a different R&D platform, or make some significant mods to the 550.
For a bunch of reasons, I decided to stay with an (albeit heavily modified) FW550.
I added arm extensions, and put on larger motors and props, which created the requisite lift capacity to ensure I could maintain vehicle control throughout the expected flight conditions.
Here's the reborn, heavily modified FW550 that ultimately became my R&D workhorse.
After spending several days (somewhat nervously) probing the flight envelope, I was able to verify I finally had a flying dev platform capably of exploring the expected flight envelope carrying the payloads I needed to carry.
The Starting Point for an R&D Copter
I suspected the flight control software process would require a lot of iterations, and flight testing was likely to include some crashes. I developed criteria for an R&D flying platform:
1) Sufficiently close in vehicle configuration that I could test the basic control theory algorithms (i.e. needed to be either a hex for an octo given what the mission vehicle was likely to be
2) Have enough physical space to allow the mounting of a variety of avionics and periperal systems
3) Have enough lift capacity and endurance to allow exploring the full mission envelope
4) Be relatively easy and cheap to repair
I ended up selecting a hobby-oriented copter, and then making a bunch of modifications to get a system I could use for development.
I started with a DJI FlameWheel 550 (FW550) . As a first step, I built a fairly stock FW550 with a Naza Flight Controller. I added an undercarriage from a RC Heli and got comfortable building and flying the platform:
I suspected the flight control software process would require a lot of iterations, and flight testing was likely to include some crashes. I developed criteria for an R&D flying platform:
1) Sufficiently close in vehicle configuration that I could test the basic control theory algorithms (i.e. needed to be either a hex for an octo given what the mission vehicle was likely to be
2) Have enough physical space to allow the mounting of a variety of avionics and periperal systems
3) Have enough lift capacity and endurance to allow exploring the full mission envelope
4) Be relatively easy and cheap to repair
I ended up selecting a hobby-oriented copter, and then making a bunch of modifications to get a system I could use for development.
I started with a DJI FlameWheel 550 (FW550) . As a first step, I built a fairly stock FW550 with a Naza Flight Controller. I added an undercarriage from a RC Heli and got comfortable building and flying the platform:
Subscribe to:
Comments (Atom)