Experiment of "MoveTo" command in Waypoint navigator on Ursa Magna 2 (December 10,  2000)

As a part of the VCL vocabulary, the "Moveto" maneuver is programmed and tested on Ursa Magna 2.
"MoveTo" is a low-speed waypoint-based maneuver with fixed heading. This maneuver is ideal for
our prestigious goal, the pursuit-evasion game.
The helicopter is guided to visit the target points with low-speed forward/backward
and sideslip maneuver while its heading is fixed. This maneuver is advantageous for low-speed
ground object tracking because the heading coordination is not required, the take-over by the ground
pilot is easier, and it minimizes the compensation of the camera coordinates. The MoveTo algorithm is
added into the waypoint navigator and first validated by Matlab Simulation and then experimented on
our Ursa Magna 2 testbed. My gratitude goes out to Hoam Chung, who helped the experiment
in the middle of his busy final examination schedule.

The maneuver demonstrated in the following movie clip is programmed by the following VCL
file:

Hover (0,0,0)rel heading=180deg duration=7sec;
MoveTo (0,-3,0)rel vel=0.5m/s heading=180deg;
MoveTo (0,3,0)rel vel=0.5m/s heading=180deg;
MoveTo (2,0,0)rel vel=0.8m/s heading=180deg;
MoveTo (-3,0,0)rel vel=0.3m/s heading=180deg;
MoveTo (4,4,0)rel vel=0.2m/s heading=180deg;
MoveTo (-1,-2,0)rel vel=0.2m/s heading=180deg;
MoveTo (-3,1,0)rel vel=0.2m/s heading=180deg;
MoveTo (3,-3,0)rel vel=0.2m/s heading=180deg;
Hover (0,0,0)rel heading=180deg duration=7sec;

Movie Clips
 
MoveTo maneuver (20MB)

Mu-synthesis controller Testing on Ursa Magna 2 (July 6,  2000)

In parallel with classical control approach, the performance of modern control theory is also investigated. Mu-Syntehsis control is a powerful linear robust control theory suitable for the control of aerodynamic systems because it can handle the structured or unstructured uncertainty of the target system. mu-synthesis control theory requires extensive knowledge of the system such as sensor noise model, uncertainty of the system on top of the MIMO system model. In this approach, the system model identified using time-domain approach is used as the base model. For uncertainty model, unstructured weigthing functions at input channels with 10% magnitude is assumed for now. The sensor noise models were synthesized based on the results of extensive study of sensor noise spectra. The resulted controller could regulate the roll and pitch angles within +- 2deg. The graph shows a little loose regulation on pitch because smaller weighting for pitch was used. The reason that only attitude regulation was attempted is because of the difficulty to obtain internally stable controller for translational velocity regulation. The expanded model for translational velocity model contains unstable zeros and the resulted mu-synthesis controller contrains unstable poles to counteract the unstable zero of the plant. Future efforts will be made on the design of velocity regulator using mu-synthesis.

Movie Clips
 
Attitude regulation with mu-synthesis controller (5MB)

Click here for for further discussion
 
 

Waypoint navigation of Ursa Magna 2 (Yamaha R-50, August 1,  2000)

Here we proudly present the flight video of Yamaha R-50 peforming waypoint navigation. This helicopter follows a flight pattern called lawn mowing. This flight pattern is very useful to sweep certain area looking for target on the ground. The attribute of waypoint including the coordinate, type, required velocity, time requirement and other options are specified using the novel concept called VCL (Vehicle Control Language). This unique interface proposed by David Shim is a human understandable script type language, generated by typing individual command or specifying them in GUI (Graphic User Interface). Then the text file is scanned for any syntax error and then the feasibility of the specified flight pattern is examined. The interpreter/executioner of the VCL is written in C and the core file can be tested as S-function format supported by  MatLAB/Simulink for any possible programming and/or logic error. After the validation in simulation, the core file, wrapper, and the controller code are compiled together and then executed. This unique framework can minimize any logistics fault error, programming mistake and system instability and safer experiments can be performed.

The maneuver demonstrated in the following movie clip is programmed by the following VCL
file:

Hover (0,0,0)rel heading=270deg duration=10sec;
FlyTo (0,-5,0)rel vel=0.5m/s stopover autoheading;
Hover (0,0,0)rel heading=0deg duration=10sec;
FlyTo (5,0,0)rel vel=0.5mps stopover autoheading;
Hover (0,0,0)rel heading=270deg duration=10sec;
FlyTo (0,-5,0)rel vel=0.5m/s stopover autoheading;
Hover (0,0,0)rel heading=180deg duration=10sec;
FlyTo (-5,0,0)rel vel=0.5mps stopover autoheading;
Hover (0,0,0)rel heading=270deg duration=10sec;
FlyTo (0,-5,0)rel vel=0.5m/s stopover autoheading;
Hover (0,0,0)rel heading=0deg duration=10sec;
FlyTo (5,0,0)rel vel=0.5mps stopover autoheading;
Hover (0,0,0)rel heading=270deg duration=10sec;

Movie Clips

 
Waypoint Navigation Part 1 (14MB)
Waypoint Navigation Part 2 (17MB)
 

 

Autonomous hovering of Ursa Magna 2 (Yamaha R-50, May 13, 2000)

After a long waiting, finally Berkeley UAV Team presents the autonomous Yamaha R-50, christened as Ursa Magna 2. Ursa Magna series is now the flag ship of the autonomous UAV fleet of BEARperforming autonomous take-off, landing, waypoint navigation. Soon they will join force with ground-based robots to perform multi-hetero agent scenario. The hovering controller is the first step towards to the aspiration. Facing 5-10 mph gusts, Ursa Magna 2 was able to hover in 80 cm accuracy.

For controller design, similar approach applied to Ursa Minor 3: system identification using prediction error method. Constant gain compensators are designed for attitude loop, velocity loop and position loop based on SISO classical control approach. While performing at acceptable level, the coupled motion still shows the need for MIMO robust controller.
Movie Clips
 
U-Magna 2 in hover-1 (5.2MB)
U-Magna 2 in hover-2 (4.1MB)
U-Magna 2 in the wind (4.0MB)
Landing (4.7MB)
 
I would like to thank Peter Ray, Ron Tal and Cedric Ma for their great help, support and advices on carrying out experiments.

Landing on a ship motion deck using Stability Augmentation System (December 4,1999)

Semi-automatic landing on a ship motion deck is perfromed using the stability augmentation system. The pilot(me) is aided by the stabilizing controls in x,y,z, and the heading. The controller can be configured to aid the full position, velocity, and heading or velocity only by a flick of a switch on the radio controller. So I fly over to the center of landing pad by velocity feedback only and then turn on the position control to land on the pad. The feedback controller makes the flight whole lot easier than having none. The landing pad was actually working while I land the helicopter a couple of times without a crash. Kudos to Tullio Celano III for his great job on the landing pad.

Movie Clips

 
Rolling motion (7.5MB)
Hovering over the landing deck (21MB)
Landing. (12.6MB)

First Full-channel autonomous hovering  (Oct. 30,1999)

Finally it's here!!! Ursa Minor 3 performed full channel(x,y,z, and yaw) hovering flight today!!!
The controller consisting of roll/pitch stabilizer, body velocity regulator and yaw angle regulator
was put to test for more than 6 minutes and the only limit was the fuel capacity. As some people with control background would guess, the velocity regulator does not perform stationary point hovering, which will be done in very near future using position estimator for DQI-NP and NovAtel GPS. Special thanks to Tullio for his help on taping this historic moment for Berkeley UAV Research. My sincere gratitude to God for His guidance.

Movie Clips
 
 
Hovering (side view, 1.5MB)
Hovering(rear view, 5.3MB)
Hovering to landing (14.7MB)