qpSWIFT: A Real-Time Sparse Quadratic Program Solver for Robotic Applications

Pandala, Abhishek; Ding, Yanran; Park, Hae Won

doi:10.1109/lra.2019.2926664

Cited by 65 publications

(30 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the previous evaluation we considered systems with a high number of states which might be disadvantageous for code generating solvers like FORCESPRO since the code size increases with the problem size [19]. Therefore, we evaluate the solvers now with a lower number of states.…”

Section: Discussionmentioning

confidence: 99%

“…Receding horizon control exposes a block diagonal structure where each time step is only coupled with the previous and the next one. IPM solvers are typically capable of exploiting this sparsity [18], [19]. This way the computational complexity of solving MPC's only grows linearly with the length of the horizon and not cubically as it would be the case if a dense QP was solved [20].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

$\mathcal{N}$IPM-MPC: An Efficient Null-Space Method Based Interior-Point Method for Model Predictive Control

Pfeiffer,

Righetti

2021

Preprint

View full text Add to dashboard Cite

Linear Model Predictive Control (MPC) is a widely used method to control systems with linear dynamics. Efficient interior-point methods have been proposed which leverage the block diagonal structure of the quadratic program (QP) resulting from the receding horizon control formulation. However, they require two matrix factorizations per interior-point method iteration, one each for the computation of the dual and the primal. Recently though an interior point method based on the null-space method has been proposed which requires only a single decomposition per iteration. While the then used nullspace basis leads to dense null-space projections, in this work we propose a sparse null-space basis which preserves the block diagonal structure of the MPC matrices. Since it is based on the inverse of the transfer matrix we introduce the notion of so-called virtual controls which enables just that invertibility. A combination of the reduced number of factorizations and omission of the evaluation of the dual lets our solver outperform others in terms of computational speed by an increasing margin dependent on the number of state and control variables.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

$\mathcal{N}$IPM-MPC: An Efficient Null-Space Method Based Interior-Point Method for Model Predictive Control

Pfeiffer,

Righetti

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The objective functions are defined as shown in ( 22), (23), and (24) below, where P˙h, P J , and P ρ are weighting matrices; B ≡ F re f 0 3×1 − Ȧv; J is a Jacobian matrix used to convert the optimization variable vd into the accelerations of joints and points; and p is an objective vector for the desired motion.…”

Section: Whole-body Control Frameworkmentioning

confidence: 99%

“…We use qpSWIFT to solve the formulated problem [23]. The solutions for vd and ρ obtained through the solver can be substituted into the floating-base dynamic equation to finally obtain the torques to be applied to the robot, as shown in (25).…”

Section: Whole-body Control Frameworkmentioning

confidence: 99%

Bipedal Walking and Impact Reduction Algorithm for a Robot with Pneumatically Driven Knees

Kim

Hong

Kim

2021

Int. J. Control Autom. Syst.

View full text Add to dashboard Cite

We propose a bipedal walking and impact reduction algorithm for a bipedal robot with pneumatically driven knees. The proposed algorithm is meaningful in that unlike in the existing studies on fully pneumatically driven robots and their control, it can overcome the low control performance of pneumatic actuators while utilizing the high compliance of pneumatic actuators through the high control performance of other joints in robots that apply pneumatic actuators only to the knee joint. Our algorithm takes advantage of whole-body control to overcome the low control performance of pneumatic systems and utilizes pneumatic compliance by simultaneously controlling the force and stiffness of the pneumatic actuators. Since a pneumatic actuator outputs a force, a force and torque converter is added to the general whole-body control framework, and a torque limit is considered as a function of the joint angle. In addition, a mechanism for selecting the stiffness of the knee is added. In particular, a force control method based on maintaining the minimum stiffness is applied to reduce the impact force when the ground conditions suddenly change. The pneumatics and the robot system were accurately modeled in a simulation, and the proposed algorithm was applied in the simulation to realize bipedal walking and to confirm the impact reduction effect in the event of a sudden ground condition change.

show abstract

“…Optimal decision strategies have been employed in applications ranging from the control of manipulators with bounded inputs [21] and magnetic microrobots under frequency constraints [22] to controlling transients in power systems [23]. ODS-based strategies belong to the larger family of optimization-based nonlinear controllers [24]- [27], whose applications in robotics are growing, thanks in part to advancements in mobile computation power.…”

mentioning

confidence: 99%

Quadratic Optimization-Based Nonlinear Control for Protein Conformation Prediction

Mohammadi¹,

Spong²

2021

Preprint

View full text Add to dashboard Cite

Predicting the final folded structure of protein molecules and simulating their folding pathways is of crucial importance for designing viral drugs and studying diseases such as Alzheimer's at the molecular level. To this end, this paper investigates the problem of protein conformation prediction under the constraint of avoiding high-entropyloss routes during folding. Using the well-established kinetostatic compliance (KCM)-based nonlinear dynamics of a protein molecule, this paper formulates the protein conformation prediction as a pointwise optimal control synthesis problem cast as a quadratic program (QP). It is shown that the KCM torques in the protein folding literature can be utilized for defining a reference vector field for the QP-based control generation problem. The resulting kinetostatic control torque inputs will be close to the KCM-based reference vector field and guaranteed to be constrained by a predetermined bound; hence, high-entropy-loss routes during folding are avoided while the energy of the molecule is decreased. I. INTRODUCTION Computer-aided prediction of the folded structure of a protein molecule lies at the heart of protein engineering, drug discovery, and investigating diseases such as Alzheimer's at the molecular and cellular levels [1]. The 3D structure of a protein molecule, known as the protein conformation, is mainly determined by its linear amino acid (AA) sequence [2]. The protein folding problem is concerned with determining the final folded structure of a protein molecule given its linear AA sequence. Studying the folding pathways is also important for designing viral drugs that cause misfolding in virus proteins [3]. Knowledge-based and physics-based methods are the two prominent computational approaches for solving the protein folding problem. Knowledge-based methods, which also include the approach of Google's DeepMind AlphaFold [4], rely on using previously determined types of folds to solve the protein folding problem [5]. In physicsbased methods, on the other hand, protein folding simulations are carried out using the first principles [6]. Physicsbased methods enjoy several advantages over their knowledge-based counterparts such as the ability to model the protein-nucleic acid interactions and to simulate protein folding pathways. However, physics-based methods relying on standard molecular dynamics (MD) suffer from numerical instabilities [7]. To address the inherent challenges of MD-based protein folding solutions, Kazerounian and collaborators [8]-[11] introduced the kinetostatic compliance

show abstract

qpSWIFT: A Real-Time Sparse Quadratic Program Solver for Robotic Applications

Cited by 65 publications

References 34 publications

$\mathcal{N}$IPM-MPC: An Efficient Null-Space Method Based Interior-Point Method for Model Predictive Control

$\mathcal{N}$IPM-MPC: An Efficient Null-Space Method Based Interior-Point Method for Model Predictive Control

Bipedal Walking and Impact Reduction Algorithm for a Robot with Pneumatically Driven Knees

Quadratic Optimization-Based Nonlinear Control for Protein Conformation Prediction

Contact Info

Product

Resources

About