Figure 1
Backpackable, rapidly-deployable bird-sized aircraft, or Micro Air Vehicles (MAVs) can quickly and easily provide soldiers and command and control personnel with an ``over-the-hill'' or ``around-the-corner'' perspective from a remote location. Furthermore, MAVs can be designed to fly in caves, tunnels, and other tight, enclosed labyrinths which are densely populated with obstacles. Previous research yielded a MAV that was capable of acquiring aerial surveillance in a 10 x 10 m2 area (about 1/3 the size of a professional basketball court). However, this much space is not always available in labyrinths. As such, the prototype required modification in order to enable flight in an area one-tenth the size.
The two most important criteria for flight within labyrinths are (a) endurance so that the aircraft can survey a significant amount of ground before the mission's end, and (b) stationary flight to ensure the MAV's safety when flying through narrow passages. Rotary-wing aircraft are obviously capable of hovering, but sacrifice endurance for this ability. Fixed-wing platforms have the endurance advantage because the lift is provided primarily by the wings as opposed to electric motors, but are unable to hover. To meet both criteria, the two advantages of each aircraft configuration were married. The resulting prototype has the form of a fixed-wing aircraft, but with an additional flight mode for hovering. This is made possible by a high thrust-to-weight ratio (T/W > 1) which allows a quick transition from cruise flight (through the stall regime of conventional fixed-wing aircraft) to hovering mode (see Fig. 2).
Back to the Table of ContentsAfter making the transition into the hovering orientation (i.e. the fuselage is vertical), the large T/W ratio enables it to hover by balancing the weight of the aircraft with the thrust from the motor. However, the aircraft is unstable in this configuration and requires an expert human pilot to constantly manipulate the aircraft's control surfaces in the following manner:
The following actions must be done in parallel (see Fig. 3)
|
 Fig. 3 |
Back to the Table of ContentsAutomating the hovering flight mode requires the aircraft attitude be measured. Microstrain's 30 gram 3DM-GX1 inertial measurement unit (IMU) outputs the MAV's orientation to an onboard processing and control unit every 10 ms (i.e. 100 Hz). After comparing the current orientation to the desired orientation to calculate the error, proportional-derivative (PD) control is implemented to yield the corresponding pulse-width modulated (PWM) elevator and rudder servo commands (see Fig. 4). A panoramic view of the flying area was desired and thus, the ailerons are not currently used to counter the effects of the motor torque. Furthermore, the throttle is controlled manually to allow for altitude adjustment. This is beneficial because surveillance at different heights (e.g. various floors, different perspectives, etc.) can be obtained.
The first autonomous hovering experiments were conducted inside an urban structure, with limited flying space, (i.e. 1 x 1 m2 area), to demonstrate the usefulness of the secondary flight mode. The aircraft was released in near-hovering orientation (i.e. the fuselage is close to vertical) and manually given enough throttle to balance the aircraft weight. The controls are simultaneously handed off to the onboard control system. Initial experiments demonstrated that the MAV was able to successfully hover in "hands-off" mode for 35 seconds (see Fig. 5). It should be noted that the aileron control surfaces remained in the neutral position (i.e. no deflection) throughout the flight. This was to purposefully allow torque roll so the MAV's bellycam could acquire panoramic footage of the flying area.
Another experiment was performed to visually contrast hovering under both manual and autonomous control. The metrics used were (1) duration of the hover before losing control and (2) stability of the aircraft while in hovering mode. The human pilot was initially given control of the aircraft and was instructed to fly around the gymnasium in cruise configuration. Then, make the transition from cruise to hover flight, and attempt to hover the aircraft for as long as possible. Fig. 6 shows two series of three video stills from the experiment, which were extracted once per second for a period of three seconds. With the plane rotating at a rate of 0.25 revolutions per second, this is enough to show two quarter rotations. It can be seen that the pilot is struggling to keep the fuselage vertical, but is able to keep the aircraft positioned over a small area (see top of Fig. 6). Out of a few trials, the human pilot was able to sustain a hover for about 30 seconds.
Next, the pilot was instructed to, again, fly in cruise configuration and manually make the transition from cruise to hover flight. However, instead of trying to hover the aircraft manually, the pilot flicked a switch on the transmitter which enabled the onboard PCU. This time, the aircraft is fixed in a vertical position and is able to hover for more than 90 seconds (see bottom of Fig. 6).
Back to the Table of ContentsAs originally thought, the torque roll did not affect the stability of the aircraft during a hover. That is, the MAV was still able to remain in the vertical position despite the rotations resulting from the motor torque. However, if this MAV was to be used in the field for surveillance and reconnaissance purposes, the view from the wireless camera onboard would have a dizzying effect as the plane was rotating at a rate of 20 rpm. Since the original aileron surface area did not create enough torque to counter the rotation when fully deflected, other alternatives had to be investigated. The first and most obvious was to increase the aileron surface area by lengthening them in the direction of the wing chord. However, this was not effective because the propeller wash during a hover only flowed over approximately 30 percent of the ailerons. Furthermore, a longer aileron when fully extended caused some airflow to completely miss the tail. This significantly effected attitude regulation during a hover.
The second approach was to mount miniature DC motors on each wingtip. The motors were positioned to produce a thrust force in opposite directions which generated a rotational force countering the motor torque (see Fig. 1). The wingtip motors are GWS EDP-20s which provide 23 grams of thrust with a 2510 direct drive propeller at 7.2 volts. With the same IMU used in the autonomous hovering experiments, the torque rolling rate (i.e. aircraft roll) can be fed back into the flight control system. Using this parameter and a setpoint roll angle of 80 degrees, PD control was implemented on the error. This determined the length of the PWM signal being output to the brushed speed controller. A schematic of the setup for autonomous hovering with torque roll regulation is shown in Fig. 8.
With the above setup, autonomous hovering experiments were conducted with and without torque roll control. The roll angle from the IMU was logged to the flight control system’s external memory and is shown in Fig. 9. The left part of Fig. 9 shows the case with no torque roll control. With the plane constantly rotating during a hover, a plot of the angular data would grow rapidly. To make each revolution more visible, the roll angle was bounded between -180 and 180 degrees. As the plane moves through one revolution, the roll angle will go from 0 to 90 degrees, 90 to 180 degrees (which is equal to -180), -180 to -90, and -90 to 0 degrees. It can be seen in Fig. 9 that the plane completes 7 full revolutions in 16 seconds, or about 26 rpm. The right portion of Fig. 9 shows the controlled condition. It can be seen that with a setpoint roll angle of 80 degrees, the plane remains at a relatively constant orientation.
Back to the Table of ContentsBack to the Table of Contents
Back to the Table of Contents