Recent technological advances in energy storage devices, sensors, actuators and information processing have boosted the development of Unmanned Aerial Vehicle (UAV) platforms with significant mission capabilities [1]. Unmanned aerial vehicles are important when it comes to perform a desired task in a dangerous and/or inaccessible environment. More recently, a growing interest in un-manned aerial vehicles (UAVs) has been shown among the research community [2]. The rotorcraft UAVs pose a set of advantages compared to the fixed wing UAVs, such as hovering, vertical takeoff and landing and an aggressive maneuvering. Within the family of the rotor-crafts, Unmanned Quadrotor Helicopters (UQHs) have gained increasing attention among scientists and engineers [3].

A quadrotor is a 4-rotor vertical take-off and landing vehicle that has the maneuvering abilities of traditional helicopters with significantly lower mechanical complexity. This low complexity increases dependability while reducing the cost of manufacturing, operation, and maintenance [4]. The basic quad-rotor design consists of four complete rotor assemblies attached at equal distances from each other and a central hub. The quad-rotor is an under-actuated, dynamic vehicle with four control inputs and six degrees of freedom [5].

All the rotors are located within the same plane and oriented such that the thrust generated by each rotor is perpendicular to the vehicle. Quadrotors use 2 sets of identical fixed pitched propellers, 2 clockwise (CW) and 2 counter-clockwise (CCW) which nullifies torque produced by rotors and eliminates the need of tail rotor used in conventional helicopters. The movement for this aircraft is achieved by manipulating the thrust produced by individual propeller.

All the rotors are located within the same plane and oriented such that the thrust generated by each rotor is perpendicular to the vehicle. Quadrotors use 2 sets of identical fixed pitched propellers, 2 clockwise (CW) and 2 counter-clockwise (CCW) which nullifies torque produced by rotors and eliminates the need of tail rotor used in conventional helicopters. The movement for this aircraft is achieved by manipulating the thrust produced by individual propeller.

Figure 1 illustrates basic flight control movement for quadrotor. Assume motor is front of aircraft and sizes of arrows represent the magnitude of thrust produced by corresponding motor. Some components, such as multi-leveled equations, graphics, and tables are not prescribed, although the various table text styles are provided. The formatter will need to create these components, incorporating the applicable criteria that follow:

• Hover or Lift/Land: To make the aircraft hover at particular altitude all four rotors are kept at same speed. To lift or land the aircraft, speed of all four rotors is increased or decreased by equal amount respectively.

• Pitch: For pitch movement around y-axis, speed of motor 1 and 3 is changed.

• Roll: For roll movement around x-axis, speed of motor 2 and 4 is changed.

• Yaw: For yaw movement around z-axis, torque difference is produced by changing speed of pair of motors on x-and y-axis.

Figure 1: Quadrotor flight control movements.

This paper is divided as follows: Section II provides the system architecture that describes the basic parts of the entire system while Section III discusses the software design and shows the flowchart of the program section IV, V show the implementation of the designed control and altitude hold. Finally, Section VI concludes the paper conclusion.

Traditional controller design

The control performs a key role in the quad-rotor’s stability, making possible to control precisely the attitude and altitude states of the aircraft. Its main goal is to make the quad-rotor moves to a new desired position (called reference) and react to external disturbances quickly and in a controlled way. Attitude control is the key element to maintain stability during flight. There were two different PID controllers that used in the experiment. The first trail was design position control only without speed control, it found that the system response was slow and not achieved the requirements so both position and speed controllers are implemented on this system to enhance its performance.

Figure 2 illustrates these four implemented controllers along with the entire system. For the pith angle, the controllers output modifies the speed of the front and rear motors, whereas in front its value is increased and in the rear its value is decreased. For the roll angle the controllers output modifies the speed of the left and right motors, whereas in the right its value is increased and in the left its value is decreased. For the yaw angle the controllers output modified the speed of the four motors, whereas in the front and rear its values are increased and in the right and left its values are decreased. The resulting rotation moment is since the pitch of the front-rear propellers is opposite to the right-left propellers. For altitude, the controllers output modifies the four motors simultaneously, being increased its values to ascend and decreased to descend the UAV [6].

Figure 2: Block diagram of the implemented PID controllers.

The modified PID controller

The industrial application used liner PID regulators due to its simple structure, good performance for several processes and tunable even without a specific model of the controlled system [7].

The traditional PID structure presents main drawbacks:

• The derivate action is calculated from the error. If the task adds a step in the reference, the output of the derivate would present an impulse. This sharp movement can saturate the actuators and push away the system from the linear zone. For this reason, most of the PID architecture presents the derivative action of the process output only.

• The desired angle to the controller is a step input or abrupt change then, because of the presence of the derivative term in the control action, the output u (t) will not involve an impulse function but will involve a sharp pulse function. Such a phenomenon is called set-point kick [8]. To avoid the set-point kick phenomenon, the operation of the derivative action will be only in the feedback path so that differentiation occurs only on the feedback signal and not on the reference signal.

The control scheme arranged in this way is called the modified PID control as shown in Figure 3.

Figure 3: Modified PI-D controller.

The implementation of the designed controllers

As shown in previous section the PID controller and Modified PID controllers were designed. In this section a PID controller and Modified PID were designed and implemented on ARDUINO board using C++ program.

PID implementation on one axe: In Figure 4, each channel was tested separately; a lot of trials were done to tune the designed PID controller. During the tuning Ziegler–Nichols second method was used.

Figure 4: Position control with disturbance.

Implementation of the modified PID on one axe: In the following test the one channel of Quadcopter is subjected it’s supposed to reject that disturbance and maintain about zero situation.

This disturbance is done by hitting the Quadcopter and observes its response while trying to maintain zero situations [9].

In Figure 5, modified PID in software design consists of two loops, inner loop which called speed controller which depends on angle’s rate and the other one called angular position which depends on angle’s value. The tuning of this control starts from the inner loop gains and then adjusts the outer loop’s gains.

Figure 5: Modified PID implementation with disturbance.

The next table shows the results of the designed PID and modified PID controllers. The experiment is conducted by going through the following steps:

• Connect the two speed controllers to the two motors and the control signal of the motors is connected to ARDUINO board.

• Connect the IMU to main ARDUINO as shown.

• Connect the power as shown to all the circuit.

• After running the quad check the rotation of the motors if there is any of them rotate in the unrequired direction then invert the side wires of the three wires of the motors.

The purpose of the designed PID controller is to maintain the one channel at zero, so the input to the PID controller is the real angles and the set point is zero position as shown in Figure 6.

Figure 6: Circuit schematic of the desired controller.

Table 1 shows the output of the designed PID controller after applying disturbance .it can be rejected but with high oscillation. The output of the designed Modified PID controller with applying disturbance can be rejected but with lower oscillation than PID.

Table 1: Results of the designed PID and modified PID controllers.

Adaptive Neuro-fuzzy Logic Controller (ANFIS) controller: In this section, as the system is nonlinear ANFIS can be used as a non-linear controller.

The input–output data of the PID controlled system at different operating points are used as training and checking data sets for ANFIS. Half of these data are used as a training data and the other half as a checking data. Training the FIS is started using hybrid optimization method after generating the initial FIS structure. In the case studied, a sixty training Epochs with zero tolerance error are used in the system training. Validation of the model obtained is performed then the generated FIS is successfully used as a controller of the system in the Simulink model. The ANFIS model structure is the same in all channels and can be seen in Figure 7. The surface viewer of the FIS that relates the inputs and the output of roll, pitch, yaw, and altitude channels can be seen in Figure 8.

Figure 7: The ANFIS model structure.

Figure 8: Shows the surface of the channels which indicate a smooth surface indicating a smooth changing from one rule to another that yields to a suitable controller.

Designed ANFIS controller hovering stabilization

The desired parameters are: [X ̇, Y ̇, Z, ψ]=[0, 0, 3, 0] where the quadcopter is to reach the altitude of 3 m and taking into consideration that there is initial error in roll and pitch 0.3 rad and yaw angles of 0.5 rad. The obtained results are shown in Figure 9.

Figure 9: Hovering stabilization using ANFIS controller.

As shown in the previous figure, ANFIS controller hovering stabilization roll response has a settling time about 7 sec also; it has a low overshoot about 0.06 rad with oscillations until reach the required value.

The pitch response has a settling time about 4 sec also; it has a high overshoot about 0.3 rad with low oscillations until reach the required value. The yaw response has a settling time about 2 sec also; it has no overshoot with low steady state error about 0.003 rad. The altitude response has a rise time about 2.7 sec also; it has no overshot with steady state error about 5 cm.

ANFIS controller hovering stabilization results are shown in Table 2.

Table 2: ANFIS controller hovering stabilization results.

Evaluation between the designed controllers

First experiment: Hovering stabilization desired [X ̇,Y ̇, Z, ψ]=[0, 0, 3, 0] the desired in this experiment is to reach the altitude of 3 m and taking into consideration that there is initial error in roll, pitch angles of 0.3 rad and yaw angle of 0.5 rad (Figure 10). In Figure 11, it can be seen that the output responses for roll obtained by modified PID controller is better than the output responses obtained by PID and ANFIS controller. There is a small oscillation and overshoot in PID and controller. There is also a high oscillation for the ANFIS. The output response of pitch obtained by modified PID controller is better than the output responses obtained by other controllers. There is a small oscillation and overshoot in both PID and ANFIS controller. The output response of yaw obtained by modified PID controller is better than the output responses obtained by other controllers. T here is an oscillation and overshoot in PID controller while steady state error with ANFIS controller. There is steady state error with ANFIS, modified PID and PID controllers for altitude channel.

Figure 10: Output responses of the desired hovering stabilization.

Figure 11: Output responses of the desired altitude hold.

The controller evaluation hovering stabilization results are shown in Table 3.

Table 3: Evaluation of designed controllers hovering stabilization.

Processor-in-loop experimental test: The following test consists in the PIL (Processor-in-Loop) that goes beyond the PC platform. This step introduces some hardware features that permit to achieve situations that are more realistic where the control algorithm will run. In PIL, the target processor is a non-real time environment and communication with external processors due to the use of specific functions installed in an integrated simulation environment installed in the host computer.

In Figure 12, PIL block is used to interface and communicate with the processor. Develop, test and validate your embedded control code on the actual processor in the model. Embedded control algorithms are tested by executing it inside a circuit simulator. This can be done by executing actual code on real hardware as shown in the following Figure 12.

Figure 12: Representative photomicrograph of histopathological features in pulmonary necropsies.

In Figure 13, PIL mode, the values calculated by the simulation tool are used as inputs for the built-in algorithm, instead of reading the actual sensors. Similarly, the output of the executed control algorithms on the processor is fed back into the simulation to drive the virtual environment. A data acquisition card is used to collect data from simulation model and send it to the embedded processor to resend it back to the simulation model. The data acquisition card is used also to overcome the problem of data transfer using serial connection in case of the two inputs. The input data to the controller in simulation is small float numbers but the serial block in MATLAB deal with unsigned integer numbers so a modified block is used as shown in following Figure 13.

Figure 13: PIL calibrated data.

The modified block consists of two stages such that bias and gain block. The output from embedded controller is calibrated to overcome the problem of data transfer.

PID-no processor in loop: Desired [X ̇, Y ̇, Z, ψ] = [0, 0, 3, 0] the desired in this experiment is to reach the altitude of 3 m and taking into consideration that there is initial error in roll, pitch angles of 0.3 rad and yaw angle of 0.5 rad.

In Figure 14, the input is send to the controller in simulation and the embedded controller in the same time, but the loop closed by using the output from simulation.

Figure 14: PID-processor in loop response.

The output response in roll has high oscillation due to high dynamic with the input to the embedded controller. The output response in pitch and yaw channels have changed due to data transfer but the output response in altitude has no change.

MOD_PID-processor in loop: Desired [X ̇,Y ̇, Z, ψ] = [0, 0, 3, 0] the desired in this experiment is to reach the altitude of 3 m and taking into consideration that there is initial error in roll, pitch angles of 0.3 rad and yaw angle of 0.5 rad. In Figure 15 the input is send to the controller in simulation and the embedded controller in the same time, but the loop closed by using the output from embedded controller.

Figure 15: Mod_PID-processor in loop response.

The output response in roll has a small change due to data transfer but the output responses in pitch, altitude and yaw have no change.

ANFIS-processor in loop: Desired [X ̇,Y ̇, Z, ψ] = [0, 0, 3, 0] the desired in this experiment is to reach the altitude of 3 m and taking into consideration that there is initial error in roll, pitch angles of 0.3 rad and yaw angle of 0.5 rad. In Figure 16 the input is send to the controller in simulation and the embedded controller in the same time, but the loop closed by using the output from embedded controller.

Figure 16: ANFIS-processor in loop response.

In Figure 17, the output responses in roll, pitch, altitude, and yaw have a change due to data transfer. This experiment shows that ANFIS has an execution time quadruple more than PID and modified PID controllers due to complex calculations.

Figure 17: Execution time.

The most important part of this paper was dedicated to finding a powerful and reliable control technique for quadrotor. The guidance and control system is designed using one of the control techniques such as modified PID, conventional PID controller and ANFIS controller. The designed controllers have the objective to circumvent some sources of uncertainty such as orientation variation. The evaluation results reveal that the designed modified PID controller has the best performance, more robust, higher stability and less control effort compared to the original and designed ANFIS controller. The modified PID has been selected for low control effort and less execution time via processor in loop experiment. This gives the green light for the next step for the implementation of modified PID controller on embedded board. Towards the verification of the designed controller, evaluation of the intended system was tested on real environment, and the real flight response was obtained and evaluated.