Online PID Self-Tuning using an Evolutionary Swarm Algorithm with Experimental Quadrotor Flight Results

Ghiglino, Pablo; Forshaw, Jason; Lappas, Vaios

doi:10.2514/6.2013-5098

Cited by 10 publications

(9 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…16 The reason for choosing this algorithm is that ABC is considered able to find PID gains that are close to the optimal solution Gao. 20 The optimization was performed for a fitness function consisting of the weighted sum of the accumulated attitude quaternion error and the accumulated torque:…”

Section: A Optimization Of Pidmentioning

confidence: 99%

“…The testbed used for the experimental testing has been used in previous research efforts by the Surrey Space Centre [15][16][17] and is shown in Figure (1). The testbed includes a Qualisys motion capture system that comprises six cameras positioned around the laboratory that track markers on a quadrotor and return data to a computer.…”

Section: A Real-life Testbedmentioning

confidence: 99%

“…However, two papers have been found that attempt to increase landing accuracy in gravity turn The simulation validation was performed based on the simulation testbed presented by Ghiglino. 16 The chosen setup consists of a gravity turn as the commanded trajectory and a vanilla UAV as the underlying controlled vehicle. The reasons for choosing this trajectory is that the gravity turn is a trajectory driven purely by attitude control, so the better the control is, the closer to the perfect gravity turn curve the UAV will be.…”

Section: A Classic Gravity Turnmentioning

confidence: 99%

See 2 more Smart Citations

Quaternion Error Based Optimal Attitude Control Applied to Pinpoint Landing

Ghiglino

Lappas

2015

AIAA Guidance, Navigation, and Control Conference

Self Cite

View full text Add to dashboard Cite

The problem of Unmanned Aerial Vehicles (UAV) trajectory tracking has been studied widely and in depth, including the application of optimal control techniques to this problem, which has also been covered frequently in literature, although mostly to simplified models of the UAV and therefore with limited range of operation. The use of more complex mathematical models for UAV control, on the other hand, is less common and practically all the available research makes use of sub-optimal control techniques. Furthermore, the application of optimal control techniques to more accurate UAV models has seldom been attempted before, and even less for quaternion attitude-based models in particular. In this paper we propose the use of two proper optimal control algorithms that rely on attitude quaternion error and its dynamics. These two control algorithms share a feature which is that although the underlying linearized model on which they are based is model-independent, the LQR cost matrices are dependent on both the particularities of the autonomous vehicle and on the UAV state, and more specifically, on the angular and linear velocities. Therefore, the tuning of these controllers becomes the tuning of cost matrices, which is a priori a simpler task than adapting a more complex controller algorithm to pinpoint landing. On the other hand, these control algorithms are based on exact linearization underlying models that, for close tracking conditions, behave effectively like linear time invariant systems. The advantage of this is two-fold. On one hand, the calculation requirements for these controllers are lighter and, more importantly, the adaption of nonlinearities specific to UAV pinpoint landing is relatively simpler. The authors of this research consider these two features to be highly useful in pinpoint landing control. NomenclatureB = Body frame [χ] BI = [φ, θ, ψ] = Euler angles from B to I δ c , δ r , δ p , δ y = Thrust, roll, pitch, yaw commands e = Error (generic) [F A ] B = [F Ax , F Ay , F Az ] = Aerodynamic forces (expressed in B) [F P ] B = [F P x , F P y , F P z ] = Propulsive forces (expressed in B) g = Gravitational acceleration due to the Earth I = Inertial frame I B Bcm B = diag {[I xx , I yy , I zz ]} = Inertial tensor of Bcm with respect to B (expressed in B) m = Mass of UAV ω = Linear control bandwidth I ω B B = [p, q, r] = Angular rate from B to I (expressed in B) [q] BI = [q 0 , q 1 , q 2 , q 3 ] = Quaternion vector from B to I r Bcm/Io I = [x, y, z] = Position of Bcm from Io (expressed in I) v I Bcm B = [u, v, w] = Velocity of Bcm with respect to I (expressed in B) [x c , y c , z c , ψ c ] = Position and yaw commanded setpoints ζ = Damping ratio 1 Control

show abstract

Section: A Optimization Of Pidmentioning

confidence: 99%

Section: A Real-life Testbedmentioning

confidence: 99%

Section: A Classic Gravity Turnmentioning

confidence: 99%

See 1 more Smart Citation

Quaternion Error Based Optimal Attitude Control Applied to Pinpoint Landing

Ghiglino

Lappas

2015

AIAA Guidance, Navigation, and Control Conference

Self Cite

View full text Add to dashboard Cite

show abstract

“…Evolutionary algorithm use has been steadily increasing in the number of published papers corresponding to an increasing number of applications over the past 20 years [1][2][3][4][5][6][7][8]. Originating as an alternative to traditional mathematical optimization techniques, the techniques now span across almost every discipline to include data compression, traveling salesmen, image processing, and more importantly for spacecraft: control theory [9][10][11][12][13], system identification [14 -19], and trajectory optimization [20][21][22][23][24][25]. Randomly searching a solution space to perform a global optimization routine can be computationally expensive and time consuming.…”

Section: Introductionmentioning

confidence: 99%

An Overview of Evolutionary Algorithms toward Spacecraft Attitude Control

Cooper¹,

Smeresky²

2020

Advances in Spacecraft Attitude Control

View full text Add to dashboard Cite

Evolutionary algorithms can be used to solve interesting problems for aeronautical and astronautical applications, and it is a must to review the fundamentals of the most common evolutionary algorithms being used for those applications. Genetic algorithms, particle swarm optimization, firefly algorithm, ant colony optimization, artificial bee colony optimization, and the cuckoo search algorithm are presented and discussed with an emphasis on astronautical applications. In summary, the genetic algorithm and its variants can be used for a large parameter space but is more efficient in global optimization using a smaller chromosome size such that the number of parameters being optimized simultaneously is less than 1000. It is found that PID controller parameters, nonlinear parameter identification, and trajectory optimization are applications ripe for the genetic algorithm. Ant colony optimization and artificial bee colony optimization are optimization routines more suited for combinatorics, such as with trajectory optimization, path planning, scheduling, and spacecraft load bearing. Particle swarm optimization, firefly algorithm, and cuckoo search algorithms are best suited for large parameter spaces due to the decrease in computation need and function calls when compared to the genetic algorithm family of optimizers. Key areas of investigation for these social evolution algorithms are in spacecraft trajectory planning and in parameter identification.

show abstract

“…In some systems, reduced-order model approximations are employed and the PID design procedures for such models are addressed by Deniz et al 9 A fundamental benefit of a PID controller is its performance in terms of minimizing tracking errors for a particular value of proportional, derivative, and integral gains depending upon the signal (step, ramp, and parabolic inputs) one intends to track. Hence, online PID self-tuning by Ghiglino et al 10 and optimal tuning of the PID controllers by Zhuang and Atherton 11 have been carried out. Online tuning is also referred to as an adaptive controller by Nishiyama et al 12 Nonoptimal techniques such as pole placement via proportional controllers are also referred as selftuning regulators by Wellstead et al, 13 Pragner and Wellstead 14 and as an adaptive pole placing controller by Elliott.…”

Section: Introductionmentioning

confidence: 99%

Proportional–integral–derivative controller family for pole placement

Ashokkumar

York

Gruber

2016

Proceedings of the Institution of Mechanical Engineers, Part C:

View full text Add to dashboard Cite

In this paper, linear time-invariant square systems are considered. A procedure to design infinitely many proportional–integral–derivative controllers, all of them assigning closed-loop poles (or closed-loop eigenvalues), at desired locations fixed in the open left half plane of the complex plane is presented. The formulation accommodates partial pole placement features. The state-space realization of the linear system incorporated with a proportional–integral–derivative controller boils down to the generalized eigenvalue problem. The generalized eigenvalue-eigenvector constraint is transformed into a system of underdetermined linear homogenous set of equations whose unknowns include proportional–integral–derivative parameters. Hence, the proportional–integral–derivative solution sets are infinitely many for the chosen closed-loop eigenvalues in the eigenvalue-eigenvector constraint. The solution set is also useful to reduce the tracking errors and improve the performance. Three examples are illustrated.

show abstract

Online PID Self-Tuning using an Evolutionary Swarm Algorithm with Experimental Quadrotor Flight Results

Cited by 10 publications

References 5 publications

Quaternion Error Based Optimal Attitude Control Applied to Pinpoint Landing

Quaternion Error Based Optimal Attitude Control Applied to Pinpoint Landing

An Overview of Evolutionary Algorithms toward Spacecraft Attitude Control

Proportional–integral–derivative controller family for pole placement

Contact Info

Product

Resources

About