We have been working for some time on a way to estimate and display important flight parameters such as angle of attack, estimated airspeed (on UAVs that do not have Pitot tubes), side slip angle or flight path vector (FPV) using a simple, robust dynamical model of the UAV and a computational method that is both fast and has computational time guarantees.
Using a hybrid systems approach, the problem boils down to finding a set of analytical solutions to least squares problems, and to identify which solution is the most likely based on an analysis of the residuals. The method executes in real time, at 50 Hz, on an APM. We tested it using hardware-in-the-loop simulation (with X-Plane). As you can see from the video, the real and estimated airspeed closely match (we added some noise to the X-Plane generated IMU data to simulate the actual performance of the APM's IMU), and the FPV/AOA/sideslip evolve realistically.
Because the dynamical model used in this algorithm is fairly robust, it requires only a small number of parameters to work. More details on our WIP paper at RTAS 2014: https://www.mpi-sws.org/~bbb/proceedings/rtas14-wip-proceedings.pdf
Comments
Such parameters are important for both control and identification.
In the first case, for example, angle-of-attack is the primary indicator of a stall. Additionally, since a basic piloting goal is to eliminate sideslip at all flying manoeuvres, you have to know it, in order to control it.
As for identification, the dynamic response of the airframe is strongly dependent upon these parameters. One could go as far as completely characterise (model) the airframe, by knowing these parameters, and, of course, an IMU.
Generally, knowing these parameters in a small UAV is hard, because you cannot mount a large airspeed probe on it, which measures them directly. Instead, you usually try to estimate these values by observing other values, such as airspeed, pitch, yaw, heading and path angle. However, this isn't reliable nor accurate, and efforts to improve on such algorithms extend to Kalman filtering, neural networks etc.
looks interesting and certainly smart . But what is this thing all about, why do we need to calculate AOA-sideslips & other chinese params, in a few words ? (for dummies like me)
Very cool job, great work.
This has been used on full scale as well, as an alternative to the traditional vane.
Hi Christian,
Nice to see this sort of work on the APM platform!
If you plan on continuing with this work then I would highly recommend you not use HIL any more. HIL adds a lot of lag, and produces a very unrealistic model of a real aircraft. I know HIL is tempting, but it just isn't good enough for this type of analytics.
Instead I'd recommend you use SITL, which provides much more accurate simulation and avoids the huge control and sensor lag that HIL has. If you fly a plane in HIL you will find you need control gains of about 20% of the gains you need for the real aircraft you are simulating. This is because in HIL the sensor and control lag means you cannot tune the plane to anything like its real performance. In SITL you can use much more normal tuning gains. SITL also has a lot more control over the simulation environment (have a look at all of the SIM_* parameters for an idea of what you can control).
The second comment I'd make is that you should try to compare the output of your system to the output of the existing estimators in the ardupilot code. For example, I'm not sure if you realise that ardupilot already has two different methods for estimating airspeed without a pitot tube. One is in the DCM code and the other is part of the new EKF code. If you disable the airspeed sensor then you will still see airspeed numbers in telemetry coming from these algorithms.
Cheers, Tridge
You will notice that the newer airliners, and all military aircraft fly AOA. Airspeed is just about irrelevant when you can instead determine AOA. Military aircraft, specifically fighters, rarely focus on airspeed for stall, and more often then not "fly angle of attack"
Most recently A5 amphib plane has introduced their own AOA system for easier pilot workload.
If aircraft were fitted with an AOA sensor/vane, programming the APM for that aircrafts specific stall would be as simple as a manual flight and a purposeful stall. The APM could recognize the nose down dove directly after a stall and record the AOA transducers position just prior to. It could then program this position as default and would be more capable of avoiding dangerous flight parameters.
-
1
-
2
-
3
of 3 Next