Home » Uncategorized
Category Archives: Uncategorized
The first flight of my 3d-printed Ducted Fan VTOL UAV was on the 27th June 2015. Usually you would not expect that a first flight work perfectly, but it did. No tuning of parameters or modifications were needed. I did attach a GoPro on the second flight to get some nice video from it. I didn’t even unpack my notebook since all the data is already stored in much higher data rate on board as it shown on the ground control station (GCS). In manual flight (commanding attitude and power to the motor) looking on the GCS is not a good idea anyway (at least for my skills). I modified the landing structure a bit to dampen a hard landing and to reduce the risk of flipping over. The following video shows the first flight with some parts of the second flight. I intentionally increased the power until I saw some movement and then accelerated fast. This minimizes the time where the vehicle is still in contact with the ground but half flying.
It was not a lucky coincidence that the vehicle did not require tuning. Some people tune and test the vehicle by holding it and see the reactions. I didn’t like that idea for multiple reasons. I would expect that it is not very precise (I have done it with quad copters before) and you need some courage depending on the vehicle to hold it with propeller rotating that close to your body. Second this vehicle makes significantly more noise than a quadcopter even though the propeller is shielded the psychological effect of the noise scared me on the beginning. Call me chicken but in a closed room that screaming noise is quite an impression (there is certainly room for aerodynamic optimisation).
So I built two test stands with some leftover wood and some 3d printed brackets with ball bearings for the suspension. One test stand was to tune the pitch parameters. Because of the symmetry along the vertical axis I just used the same parameters for roll (in hovering mode). For the yaw parameters I built another test stand where the vehicle was suspended along its vertical axis. The tuning of the parameters was relatively straight forward as well as writing the necessary mixer. Additionally I added some wool thread on the the stators, control surfaces and the inlet of the duct to see if there some airflow separation occurs. On the test stands I didn’t see any airflow separations, even though the incoming airflow was very gusty, since I did those tests in a relatively small room. One wool thread showed the back flow of the air between the propeller and the inside wall of the duct very nicely. The gap , 1mm is relatively large and can be reduced if the inside wall of the duct could be stiffened by an additional rib at this area.
I between the different test stands a few times to make sure that parameters aren’t having any unwanted influence onto other axis. That gave me also a feeling how far I can go with the parameter tuning and what performance I could expect. Once I was satisfied with the parameter tuning I modified the landing structure and waited for a day with sunshine and calm winds (which is not that often the case here). The last thing I did before I left for the first flight was to measure the weight. You never know it might be much more difficult to figure out what the weight of the vehicle was after the first flight. It is 1.3kg with battery ready for take-off.
One last picture before the take-off. Yes, I used in that case a peace of duct tape to hold the receiver in place.
The first flight was very pleasant. I could have not expected more. For the second flight I mounted a GoPro camera on the vehicle and tested the reactions on my inputs. I never used full power, since the vehicle accelerated vertically very rapid with estimated ¾ power.
The next steps will be to tune the parameters for position hold and way-point following (in hovering mode). For the transition into horizontal flight I will have to do some add-ons. Most likely I will build another airframe since the costs for the airframe are insignificant and I can in-cooperate all the changes to improve the assembly and the vehicle it self.
At this point I want to thank all the developers of the PX4 project, without their work I would need a few more weeks and would still have just a basic flight controller. I love that they integrated the off-board control mode, which makes it super easy to send commands and receive vehicle information form a small Linux computer. I am using again the BeagleBone Black (this time just for the high-level control and digital video transmission).
A 3D printed UAV prototype which makes use of the advantages of 3D printing
Though Ducted Fans UAVs as of now are not very common, my personal opinion is that this will change in the future, since they have certain advantages over other concepts. Just one single engine is necessary which makes the use of internal combustion engine more viable than for multi-rotors. Also a transition into a very efficient horizontal flight is possible (with modifications). This is a project which started in 2001 as a Ducted Fan VTOL in a configuration where the centre of gravity is located beneath trust vectoring. Since there were no hobby autopilots available at the time, it was only equipped with one gyroscope and a mechanical mixer to control the vehicle around the vertical axis. The vehicle was destroyed on the first flight with minor success.
With the uprising of small accessible autopilots and increasing knowledge, the concept of this configuration was given up in favour to a concept with the centre of gravity above the control surfaces. The lack of time and the high effort to create relative complex moulds for composite sandwich parts and the parts themselves slowed down the progress. A new upcoming technology (3D printing) changed this project again. 3D printing is perfect for complex shaped parts in low quantity. It is perfect for creating objects with surfaces curved in two directions (like on a duct). For a prototypes where the minimal weight is not primary, 3D printing might be the best solution such as with this project. I modified my CAD model of the Ducted Fan VTOL UAV for 3D printed plastic parts. I also built a 3D printer which fits my purpose with a built volume of 40cm diameter and 40cm hight.
3D printing has further an advantage that mounting brackets and other small parts can be included in the structure. In conventional building methods they are usually added in a later stage of fabrication. This reduces the time for mounting servos and other parts which can be a very time consuming task. In this prototype mountings for servos are already included in the structure. The control surfaces are directly connected to the servos and extremely light. Most of the structure consists of 0.35 mm (1/72”) to 0.5 mm (1/48”) thick walls with rips to stiffen it out.
Originally I had planned to equip the UAV with a self built autopilot. This autopilot is based on a Beagelbone as a processing unit connected via (real-time) Ethernet to a micro-controller which serves as an real-time interface between the sensors and actuators. This has the advantage that the processing unit can be exchanged according to the requirements by any other computer with a Ethernet port. The disadvantage is that compared to other available open source systems the developed system is in a very immature stadium. So I decided to use an autopilot based on the PX4 which I recently used for other projects and has a huge community.
The UAV is powered by one 3 cell LiPo battery with 3500mAh. The outside diameter of the duct is 30cm (12”), the height is 55cm (22”) and the take-off weight is about 1.2kg (2.6 pound). It has not yet performed its first flight, but some bench tests have been conducted. The produced trust is sufficient, but no measurements have been taken so far. Also the stators together with the control surfaces do a great job in elimination the reaction forces form the motor. Wool threads were fitted to the stators inside the duct to see if any airflow separations occur, which would inevitably cause a loss of control. The same will be done for the inlet to see if any separation occur trough influences like wind gusts or extreme manoeuvres.
Once the missing parts arrive the next step will be a free test flight equipped with analysis equipment (threads on the inlet, outlet and the stators recorded by multiple cameras). Together with the recording function of the autopilot this should maximize the information gain to improve any non perfect test flight. This will hopefully happen soon and I will give an update.
Personally I can’t imagine doing that kind of work anymore without a 3D printer. Even though the parts might be not perfect with a home made printer and composite materials might be more suitable, 3d printed (prototype) parts can greatly improve the manufacturing efficiency and let you concentrate on the task you really want to achieve. In my case it is to bring this Ducted Fan UAV into the air and modify it that way that it can be used as part of a fully automatic Unmanned Aerial System.
By fully automatic I mean a system which consists of a box which has a power- and internet connection which opens on command, the UAV takes off, fulfils its mission (possibly makes a transition into a efficient horizontal flight) lands fully automatic into the box ,and gets recharged or refuelled. This box can be located on the top of any building or tower or on a boat/vehicle. The user exclusively concentrates on the mission. Currently I am looking for people who share similar interests for cooperation or collaboration. If you are interested or you know somebody who could be interested then feel free to contact me at: info (at) arminstrobel.com
The basic concept of this airborne launch and recovery system is based on the the commonly used In Flight Refuelling (IFR) system via probe and drogue. Instead of hooking up a fuel nozzle the whole UAV is hooked into the drogue, and with a winch it is pulled in a pod under the wing where the UAV is stored while not in use or being refuelled / recharged. For every UAV one pod dedicated, except when a pod is mounted in the payload bay of the mother plane. Multiples of pods can be mounted on existing stations.
Since the properties and size of the drogue are similar to the UAV both react similar to turbulences. In this suggested approach the coupling of the UAV with the drogue is outside of the area of the propeller wash and the down wash of the wing and other disturbances. Further this allows to define a safety zone around the mother airplane to prevent any threat from the UAV to the mother airplane or its crew.
In the example of the C-130 one pod can be mounted inside the fuselage accessible for the crew. Through opening the back door the pod can be brought into position to deploy the UAV. This gives the ability to deploy and store multiple UAVs with one pod and allows access for the flight crew during flight to the UAV and its payload.
The rough closing in for recovery is done by transmitting the position, speed and heading of the mother airplane via an encrypted data link. The UAV uses this information to calculate a possible rendezvous point and adjusts its flight path.
If the pilot of the mother airplane is keeping its heading and speed, the rendezvous manoeuvre is straight forward. To give the mother airplane pilot more flexibility a predictive probability based approach can be used to control the UAV in rough closing in phase. This is very similar to a submitted publication of mine (Predictive Probability Based Collision Avoidance for Unmanned Aerial Vehicles).
When the UAV is close enough a precision guidance system is used to couple. The precision guidance system persists of a infra red (IR) light source (preferable wave length 940nm) mounted in the middle of the drogue and a corresponding IR camera with a high update rate in the probe at the head of the UAV. This allows the UAV to close precisely the probe in the drogue where a coupling mechanism fixes it (Figure 2). The winch then pulls the drogue with the coupled UAV into the pod where the UAV is stored and refuelled for the next deploy (Figure 3).
The UAV can be either specially designed for this kind of application or can be refitted with the probe. The probe can interoperate all necessary systems needed for deploy and recovery. The adaptation of a different UAV can be done by integrating the probe and adjusting the shape of the storage bay in the pod for the corresponding UAV.
To deploy the UAV the winch can either unwind and at a distance release the UAV or it can be directly released from the pod. The pods position and the drag of the UAV prevent any contact with the mother airplane. For a safe deploy even under imperfect conditions or during a dive or deceleration additional drag can be generated by releasing a drag chute prior to deployment.
The necessary systems for this type of in flight deployment and recovery of UAVs includes following Systems:
- The UAV
- Airframe with payload, propulsion and control for normal operations
- High level control for distributed control (coordination)
- IR-Sensor with control for recovery
- Control station for distributed coordination of several Vehicles
- The launch and recovery system (including storage and refuelling)
- Storage bay
- Refuelling / recharging system
- Display for the Pilot to visualize the area he should be or stay for recovery or deploying
Figure 5 shows an overview of those systems and there relations to each other.
The drogue has two proposes. It is supposed to keep the latching hole stable in the air so that the UAV can home into it by using the IR light source inside the latching hole (Figure 6). The second purpose is to guide the pin into the latching hole, where the pin is latched. The IR light source is either supplied by electric wires in the rope or by a small battery which gets inductive recharged when it is in storage position.
The probe is designed to house all additional systems for the UAV. An IR camera behind an optical notch filter window is located on the front of the probe (Figure 7). Behind that camera is the control system with the data link installed. These are surrounded by the latching mechanism. The whole probe is supposed to slide into the latching hole of the drogue. The probe’s power is supplied by the UAV and gets information such as position and speed from the autopilot. This information is used together with the information of the mother airplane via data link to line up the UAV short behind the mother airplane. The precision IR based homing system is used for the final phase. In both phases the control system sends correction data to the UAV’s autopilot.
The pod is equipped with the winch, a data link, refuel system and the control system for it. The rear of the pod contains the storage bay for the UAV. The back of the storage bay is shaped in a way that the drogue and the the UAV guides easy into the storage bay. In storage position (Figure 10) the UAV is held only by the locking mechanism in the drogue and the storage bay shape. The Figure 8 to 10 shows the pulling of the UAV into the storage position.
Mother Airplane Pilot Display
This display gives the pilot of the mother airplane the information regarding where the airplane has to be to be able to recover the UAV in a certain time period. The different time periods can be displayed with different colors. The stand off areas for the mother airplane can also be integrated so the pilot is able to determine where they are and where they should be with just one look.
Distributed Control of the UAVs
The UAVs can be either operated as independent vehicles with different tasks allocated by the operator or as a small swarm. In the first implementation multiple operators will control the UAVs and coordinate their mission.
In a further development stage a coordinated swarm control may be beneficial depending on the mission. In my doctoral thesis (Unmanned Aerial Vehicle Swarming with Distributed Nonlinear Model-Predictive Control) one possible approach of swarm coordination and control is described.
To control a swarm of Unmanned Aerial Vehicles (UAVs) efficiently, every UAV has to have a higher level of autonomy than just a single UAV. The operator does not want and can not give every single member of a swarm detailed commands. The autonomy has to ensure that the operator can give high level commands and every member of the swarm is collaborative and reliable fulfilling its desired aim.
Applications and Use cases
This concept, of launching and recovering, can be applied for every UAV which can be equipped with the probe and its flight envelope is a sufficient match to the mother airplane. This includes fast propeller driven UAVs in pusher and twin engine configurations as well as all turbine driven UAVs. The UAV’s maximum speed needs to be sufficient above the minimum speed for safe operations of the mother airplane. In some cases the UAV wing and tail structure (which are exposed in storage) may need to be modified to ensure the survivability in storage position for the complete speed envelope of the mother airplane.
The use of a vertical take off and land (VTOL) UAV like a modified Aerovel Flexrotor UAV with higher maximum speed would allow to deploy precisely and recover small payloads to and from everywhere. This will offer new kinds of operations and missions.
Advantages / Disadvantages
- Based on existing technologies like In Flight Refuelling and glider towing.
- Mainly use of commercial of the shelf components possible.
- Coupling of the UAV can be done far behind the mother airplane, which ensures a safety buffer.
- Drogue can be designed in a way that the reaction on turbulences are very similar to those of the UAV. This will allow a coupling also under imperfect conditions.
- Fast recovery, refuelling / recharging and redeploy possible.
- Adaptable for many existing UAV / UAS.
- Modular design.
- Only one UAV can be recovered per pod if they are mounted on the wing and not inside of the fuselage such as in Figure 4.
After I had done all the research for my doctoral thesis, I started the not fun part of my thesis (the writing of the actual document). So I decided that I need a fun project on the side to relax. During that time I was in Germany where I had access to a workshop. To make use of the workshop and to accelerate a long time ongoing project I decided to built a 3D printer.
There are a huge amount of different designs available and the community does an amazing job to design and supply all necessary parts. The commercially available printer appeared me too flimsy for there money. Also the fact that with the open source/hardware community almost anybody can build relative easy there own printer I decided to start this project.
My choice was a delta type printer with a built volume of 40cm x 40cm x 40cm. Since I eventually wanted to be able to use the printer to mill moulds out of high density foam (for fabrication of lightweight composite parts), I designed the printer much stiffer. I also used some more expensive linear bearings. With this kinematics the rotational loads on the rails are minimal. So I was able to used uses just one linear slide per axis. This reduces the cost significantly even I had to used the more expensive linear bearings.
With a drill press, a metal hacksaw, a vice, some tabs and some drill bits I fabricated the few parts which are not off the shelf. I took those parts with me to Canada. Obviously my luggage got multiple times carefully screened by the border authorities and I had some really nice conversations.
In Canada I fabricated the rather large wood parts and started to assemble all the parts. After multiple modification and adjustments I was finally able to print parts with acceptable quality. Thanks to the open source community the motivations and adjustments where much faster than it would probably have been without those information.
After several smaller test parts I was finally able to print the large part, for which the printer was originally built. It is duct, much more I will not reveal for now. The quality of this part is acceptable, but can most likely and will be hopefully soon improved by a stronger (geared) extruder stepper motor, a heated chamber/bed and other features.
The printer in action (time lapse):
The printer in action (real time):