Autonomous rotorcraft provide improved capability in performing military missions such as reconnaissance, targeting, border patrol, and environmental sensing. A common difficulty in applying miniature rotorcraft to these areas is the complexity and specialization of the control. In general, rotorcraft have extreme vibration that make miniature inertial measurement difficult. Sources of vibration include the main rotor, tail rotor, and blade flapping dynamics. Typical sensors include MEMS accelerometers, which are sensitive to vibration. Inclusion of alternative and or redundant sensors may be used to reduce vibration sensitivity and add useful additional feedback.
The estimation of orientation using multiple vector measurements is a standard attitude determination problem that can be separated into two types of solutions: deterministic and optimal algorithms. Deterministic algorithms use two measurement vectors (four pieces of information) and discard part of the measurements so that the orientation parameters can be found. A common deterministic algorithm is the TRIAD algorithm (TRI-axial Attitude Determination system), which uses two measurements, the first assumed to be more reliable than the second. Drawbacks of the TRIAD algorithm are that it accommodates only two observations, and some accuracy is lost because part of the second measurement is discarded.
Optimal attitude determination algorithms differ from deterministic algorithms by computing a best estimate that minimizes a loss function J. A solution to the optimal attitude determination problem was previously developed using a quaternion representation that leads to an eigenvalue equation solution called the q-method. An efficient approximate method to the q-method was later developed called QUEST that allows the approximation of the optimal quaternion without solving the eigenvalue problem.
The work reported here evaluates the well-known attitude determination algorithms for use on small rotorcraft. In order to reduce noise from vibration, two alternative methods are proposed and evaluated. The first alternative algorithm adds into QUEST a pseudo-measurement vector derived from rotational kinematics. As a comparison, a simple gyro-compensated tilt sensor and compass was also developed. Finally, the orientation algorithms were tested on a small autonomous hovering rotorcraft.
The T-Rex 600 CF helicopter was chosen as the test platform. It is a full-size radio-controlled helicopter consisting of carbon fiber (CF) and CNC aluminum parts. The helicopter measures almost 4 feet in length, uses full-size 600-mm rotor blades, and is powered by a ballistic combination of both a large brushless motor and lithium packs (in series). The head uses mixing in the transmitter to drive three servos linked to the three stationary swash plate balls at 120-degree intervals. This moves the swash plate up and down for collective pitch and tilts it for cyclic control by moving the servos in combination. Because these rotors produce small damping moments in comparison to larger helicopters, the design features stabilizer bars for ease of handling.
The T-Rex 600 is equipped with a mechanical Bell-Hiller stabilizer bar that effectively applies lagged rate feedback to the two cyclic control channels. This system may be regarded as a secondary rotor attached to the shaft below the main rotor by an unrestrained teetering hinge. Aerodynamic paddles are attached to the ends of a rod. Cyclic pitch and roll are inputs transmitted to the stabilizer bar. But unlike the main rotor, the stabilizer bar has no collective.
The small rotorcraft control law transforms the task of controlling the rotorcraft over the flight into one of controlling the servo applications for roll, pitch, yaw, and altitude. Information about the desired hover point is used to determine when to correctly adjust the swash plate and the heading. In hover, the desired altitude simply becomes the desired height above the ground. Desired roll and pitch are related to the position of the helicopter.
The flight control system hardware includes multiple microprocessors, GPS, three- axis accelerometers, three- axis magnetometer, three-axis gyroscope, electronic variometer, airspeed sensor, and barometric altimeter. The gyro-compensated tilt sensor and compass was selected as the orientation algorithm due to its comparable performance with the optimal QUEST solution and its simplicity.
The ability to communicate wirelessly with the autopilot is a critical component in the autonomous flight development. This achievement is made possible by the graphical user interface (GUI) that allows the user to view real-time sensor data, change desired wave points in current flight, store proportional integral-derivative gains, calibrate onboard sensors and servo parameters, and view selected variables of the control algorithm.
For the flight test, the T-Rex 600 implemented with the autopilot was placed stationary on the field until a reasonable GPS accuracy was reached. While the helicopter was stationary, all sensors were checked and calibrated appropriately. The helicopter was then placed in the center of the field and that specific GPS point was entered as the desired wave point into the graphical interface. With the complete control algorithm programmed and all servos attached to the autopilot, the helicopter was taken to an altitude of 5 feet manually and switched to autonomous mode. A camera was mounted to the bottom of the landing gear to track the desired point in the center of the field, which was marked by a 2 × 2 × 2-foot case.
Sensor noise that is amplified by the PID gains caused the helicopter to oscillate about the desired height of 30 feet by ±10 feet. The helicopter speed remained about 3 feet/second for the duration of the flight. The orientation of the rotorcraft tracked the desired angles appropriately.
This work was done by Nathan Slegers of the University of Alabama- Huntsville for the Army Research Office.
This Brief includes a Technical Support Package (TSP).
Miniature Rotorcraft Flight Control Stabilization System
(reference ARL-0048) is currently available for download from the TSP library.
Don't have an account? Sign up here.