Nonlinear model predictive control for aerial manipulation
read more
Citations
Data-Driven MPC for Quadrotors
Fractional-order sliding mode control of uncertain QUAVs with time-varying state constraints
Model Predictive Control for Micro Aerial Vehicles: A Survey.
Inspection-while-flying: An autonomous contact-based nondestructive test using UAV-tools
Nonlinear Model Predictive Guidance for Fixed-wing UAVs Using Identified Control Augmented Dynamics
References
Survey Constrained model predictive control: Stability and optimality
Minimum snap trajectory generation and control for quadrotors
ACADO toolkit—An open-source framework for automatic control and dynamic optimization
Singularity-robust task-priority redundancy resolution for real-time kinematic control of robot manipulators
Control of an aerial robot with multi-link arm for assembly tasks
Related Papers (5)
Frequently Asked Questions (14)
Q2. What is the definition of a multirotor?
At a high level of control, a multirotor is an underactuated vehicle with only 4 DoFs, i.e., the platform tilting (roll and pitch variables) is used by the low level attitude controller to produce desired translational velocities of the vehicle.
Q3. What is the definition of a NMPC?
NMPC refers to particular MPC problems where the process model is nonlinear, the cost functional is nonquadratic or general nonlinear constraints are used.
Q4. What is the disadvantage of a weighting strategy?
A clear disadvantage of a weighting strategy is the lack of good solutions when tasks are antagonistic (i.e., when the accomplishment of a task requires the oposite solution of an other task).
Q5. What is the cost function for the arm CoG?
In order to minimize this actuation effort and avoid instability, it is beneficial to design a task to favor this alignment, such ash2 = N−1∑ i=0 ||bpGxy(xk+i)||2W2 + || bpGxy (xk+N ) ||2Ws2 ,(10) where bpGxy(xk+i), is the vector describing the position of the arm CoG projected onto the xy plane of the body reference frame during the prediction horizon.
Q6. What is the cost function for a multirotor?
In a number of works, collected in [23], it is pointed out that terminal cost weights are a key ingredient to achieve stability with NMPCs.
Q7. What is the eigenvalue of the matrix Je,qJ>e?
Notice that, far from singularity points, the matrix Je,qJ>e,q is symmetric positive definite, then its eigenvalues are real and positive.
Q8. What are the velocity bounds of the controller?
On the other hand, velocity bounds correspond to control action constraints, thus they can be obtained by setting values of umin and umax.
Q9. What are the tasks that are presented in Fig. 8?
Among them, tasks to track trajectories with the arm end effector, improve the arm manipulability or align the arm CoG with the platform gravitational vector have been discussed.
Q10. What is the optimal control problem at time tk?
Considering a generic dynamic system, which has a state x and is controlled by the variables in u, the solution of the optimal control problem at time tk = kTs (∀ 1 < k < N ), where
Q11. What is the cost function for the secondary task?
For a secondary task, the minimization of the cost functional follows the same procedure as for the primary task, but this time incorporating the constraint resulting from solving the primary task (see (3)), defined byg1(xk,u, tk) = 0 , ∀k.
Q12. What is the effect of adding a secondary task to the UAM?
In these particular simulations the authors removed all motion bounds to show how, in the case where only the tracking task is present (Fig. 6(a)), the platform and end effector are moving in similar vertical positions, thus with the arm fully extended in a forward position.
Q13. What is the cost function for each time step?
Drawing inspiration from [20], the authors propose to compute a cascade of optimizations for each time step in order to minimize different cost functionals, each one related to a different task.
Q14. What is the weight of the i-th joint?
To do so, the authors impose a safety distance between the joints and the aerial base withbzi (xk) ≥ 0.1 [m] bze (xk) ≥ 0.1 [m] ,(17)where bze is the position of the end effector in the z direction of the body reference frame and, similarly, bzi is the distance of the i-th joint.