An intelligent landing system based on a human skill model

Longitudinal auto‐landing controller design via adaptive backstepping

“…Wind disturbances are modeled by a Dryden spectrum [6,18]. The disturbance equations for the longitudinal wind and the vertical wind are represented by…”

“…Detailed descriptions of the conventional PI controller can be found in [5,6,17], including airspeed and elevator control. The throttle command to maintain aircraft's airspeed at a constant value is generated by…”

“…To show the effectiveness of the proposed controller, all simulations following the auto-landing commands should meet the following conditions and scenarios: (i) the time constant ( ) and the flare initiation height (h 0 ) must be pre-determined in Table I, and the values of and h 0 are 2.98 s and 18.47 ft, respectively; (ii) the UAV is trimmed at a speed of 70 knots/altitude of 200 ft and the initial distance is 4000 ft away from touchdown point; (iii) the glide-slope mode is engaged from 3816 ft (H 0 / tan( GS )) away from touchdown point depending on the initial altitude (H = 200 ft) and the desired glide-slope angle; (iv) the elevator actuator is modeled as a first-order system model with a time constant of 0.1 s with a rate limit of 60 • /s, and the throttle servo is ignored; (v) the parameters K 1 , K 2 , and K 3 are easily obtained from (49), and K 1 = 1.5, K 2 = 2.3, and K 3 = 8.0 are selected in the simulations with j ( j= p,i, ,q, f,v,a,d) = 20; (vi) although tracking flight path angle is suitable for varying descent line by using a different glide-slope angle before final approach, a single glide-slope trajectory is used in all simulations; (vii) the successful landing specifications are given in Table II. Note that these specifications are referred to [6,15,33]. Iiguni and Akiyoshi [6] offered the specifications of successful touchdown for transport aircraft, Shue and Agarwal [15] proposed some constrains of the glide-slope maneuver and the bound of the flare maneuver at the landing angle, and Pashikar et al [33] presented the Pill Box conditions for touchdown; any deviation from these constrains, the aircraft will be request to abort the landing task.…”

Neuro-aided H2 Controller Design for Aircraft under Actuator Failure

“…Although the auto-landing controller is still PID, this paper explores the use of neural network in auto-landing control system by incorporating it into the inner loop to generate the desired pitch angle command. Further, in [7], a neural network is used as an auto-landing controller, where a feed-forward network is trained off-line to replace the original PID controller. In [8], Napolitano and Kincheloe present the investigation of online learning neural controllers in the context of the aircraft autopilot functions and of stability augmentation systems for both longitudinal and lateral directional dynamics.…”

Abstraction and Implementation of Human Skill by Hybrid Dynamical System Theory-Application to an Automatic Driving System-

“…For example, in the field of aerial navigation, an automatic landing system has been developed on the basis of a human skill model, and the model is expressed as a nonlinear I/O mapping from the aircraft state to the control command provided by a human expert (1) . For robot balance control, a human control strategy model has been trained to abstract the operator's skill in controlling the tilt-up motion, and then a human-based controller has been developed such that the robot can recover from fall automatically (2) .…”