The wheeled and legged robots we have discussed in the previous sections were designed to work on land, but there are also robots that fly through the air (e.g.,Montgomery, Fagg, and Bekey 1995) or swim under water (e.g., Yuh, Ura, and Bekey 1997). It shows an autonomous robot helicopter, AVATAR (Autonomous Vehicle Aerial Tracking and Reconnaissance), developed at the University of Southern California (USC) by the author’s colleagues and their students.
Clearly, the problems inherent in the control of a robotic air vehicle are quite different from those involved in the control of land vehicles. First, to remain airborne,the vehicle must generate suficient lift to overcome both drag and gravitational forces. This implies a need for suficient forward speed for an aircraft and suficient rotational velocity for the rotor (e¤ectively, a rotary wing) on a helicopter. Speed, in turn, means that dynamic e¤ects cannot be neglected, as they often can be and are with relatively slow land vehicles.
The dynamics of helicopters are quite complex, since they include aerodynamics,blade bending (and possible oscillations), and the interaction among various control modes. Much of the control in helicopters is obtained by adjustment of the pitch of the rotor blades, once every revolution (hence referred to as ‘‘cyclic’’). This cyclic adjustment increases lift on one blade while decreasing it on the other, affecting both the vehicle’s pitch and its roll. A control mode called the ‘‘collective’’ changes the pitch of the rotor blades by the same amount (collectively); this change a¤ects the thrust, thus increasing or decreasing the helicopter’s lift. Tail rotor pitch affects the vehicle’s yaw. There is a great deal of cross-coupling between control modes.
For example, changes in the thrust level (from the throttle or the collective) produce torques about the yaw axis, which need to be counteracted by the tail rotor to ensure that the helicopter’s heading does not change. Hence, the resulting differential equations governing helicopter dynamics are highly nonlinear. For this reason many investigators in the area of robot helicopters (e.g., Montgomery [1999]) have not attempted to solve the complete set of equations governing the dynamics, relying rather on heuristics, simulation, and trial-and-error methods to find proper values for control system gains.
