Search

Learning-based Nonlinear Model Predictive Control (LbNMPC) is a promising framework for applying NMPC using dynamics models obtained directly from riding data. We developed a LbNMPC controller based on a black-box dynamics model and compare it with a previously published NMPC controller based on a physics-based description. The controllers have been tested in a high-fidelity simulation framework, and the LbNMPC outperforms the nominal one in terms of tracking indexes, while requiring twice the solution time.

This paper proposes an efficient QP solver for electric motors.

This paper proposes a fast and accurate solver for implicit Continuous Set Model Predictive Control for the current control loop of synchronous motor drives with input constraints, allowing for reaching the maximum voltage feasible set. The related control problem requires an iterative solver to find the optimal solution. The real-time certification of the algorithm is of paramount importance to move the technology toward industrial-scale applications.A relevant feature of the proposed solver is that the total number of operations can be computed in the worst-case scenario. Thus, the maximum computational time is known a priori. The solver is deeply illustrated, showing its feasibility for real-time applications in the microseconds range by means of experimental tests.The proposed method outperforms general-purpose algorithms in terms of computation time, while keeping the same accuracy.

This paper proposes an effective solver for implicit Continuous Set Model Predictive Control for the current control loop of synchronous motor drives with input constraints, allowing for reaching the maximum voltage feasible set. The related quadratic programming problem requires an iterative solver to find the optimal solution. The real-time certification of the algorithm is of paramount importance to move the technology toward industrial-scale applications by the proposed solver. The total number of operations can be computed in the worst-case scenario, thus the maximum computational time is known a priori. The solver is deeply illustrated, showing its feasibility for real-time applications in the microseconds range by means of experimental tests. Promising results are obtained with respect to well known general purpose solvers.

We propose a non-linear model predictive scheme for planning fuel efficient maneuvers of small spacecrafts that shall rendezvous space debris. The paper addresses the specific issues of potential limited on-board computational capabilities and low-thrust actuators in the chasing spacecraft, and solves them by using a novel MatLab-based toolbox for real-time non-linear model predictive control (MPC) called MATMPC. This tool computes the MPC rendezvous maneuvering solution in a numerically efficient way, and this allows to greatly extend the prediction horizon length. This implies that the overall MPC scheme can compute solutions that account for the long time-scales that usually characterize the low-thrust rendezvous maneuvers. The so-developed controller is then tested in a realistic scenario that includes all the near-Earth environmental disturbances. We thus show, through numerical simulations, that this MPC method can successfully be used to perform a fuel-efficient rendezvous maneuver with an uncontrolled object, plus evaluate performance indexes such as mission duration, fuel consumption, and robustness against sensor and process noises © 2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/

IEEE Robotics and Automation Letters, 6 (3), ISSN:2377-3766

In dynamic driving simulators, the experience of operating a vehicle is reproduced by combining visual stimuli generated by graphical rendering with inertial stimuli generated by platform motion. Due to inherent limitations of the platform workspace, inertial stimulation is subject to shortcomings in the form of missing cues, false cues, and/or scaling errors, which negatively affect simulation fidelity. In the present study, we aim at quantifying the relative contribution of an active somatosensory stimulation to the perceived intensity of self-motion, relative to other sensory systems. Participants judged the intensity of longitudinal and lateral driving maneuvers in a dynamic driving simulator in passive driving conditions, with and without additional active somatosensory stimulation, as provided by an Active Seat (AS) and Active Belts (AB) integrated system (ASB). The results show that ASB enhances the perceived intensity of sustained decelerations, and increases the precision of acceleration perception overall. Our findings are consistent with models of perception, and indicate that active somatosensory stimulation can indeed be used to improve simulation fidelity.

Vehicle attitude estimation is nowadays essential for a wide range of applications, e.g. guidance of unmanned vehicles, robotics and automotive controls. In this paper, the attitude estimation problem is solved by means of a velocity-aided, Extended Kalman Filter with correlated noise (CEKF), exploiting the intrinsic correlation between sensor noise in a velocity-aided model. Reconstruction of a motorcycle attitude is considered as a use case. The problem is addressed in a simulation scenario under noisy measurements. Performance of the CEKF is compared to that of the classic EKF formulation by evaluating the RMS of the reconstruction error.

Move blocking (MB) is a widely used strategy to reduce the degrees of freedom (DoFs) of the optimal control problem (OCP) arising in receding horizon control. The size of the OCP is reduced by forcing the input variables to be constant over multiple discretisation steps. In this study, the authors focus on developing computationally efficient MB schemes for multiple shooting-based non-linear model predictive control (NMPC). The DoFs of the OCP is reduced by introducing MB in the shooting step, resulting in a smaller but sparse OCP. Therefore, the discretisation accuracy and level of sparsity are maintained. A condensing algorithm that exploits the sparsity structure of the OCP is proposed, that allows to reduce the computation complexity of condensing from quadratic to linear in the number of discretisation nodes. As a result, active-set methods with warm-start strategy can be efficiently employed, thus allowing the use of a longer prediction horizon. A detailed comparison between the proposed scheme and the non-uniform grid NMPC is given. Effectiveness of the algorithm in reducing computational burden while maintaining optimisation accuracy and constraints fulfilment is shown by means of simulations with two different problems.

Virtual prototyping tools are widely used in the development of new motorcycles, but the instability of two-wheel vehicles leads to the necessity of implementing a vehicle driver. In this brief, the development of an effective virtual rider (VR) is described that is capable of controlling the motorcycle during complex, high performance maneuvers. The control inputs are the steering angle, the lateral rider movement, and the throttle/braking effort. The proposed motorcycle model consists of a plane that can roll and slide (both in $x$ - and $y$ -directions), on which a moving point mass with lateral degree of freedom (DoF) is attached, representing the rider. The controller is designed within the nonlinear model predictive control (MPC) framework. The implementation is developed in MATMPC, an open-source toolbox, and tested in co-simulation with the commercial software VI-BikeRealTime , specifically designed to reproduce vehicles behavior with high fidelity. A chicane and a lap on a challenging track have been used to both show flexibility and evaluate performance, demonstrating the capability of the proposed VR of handling complex maneuvers.

More and more sophisticated assisted/autonomous vehicles are becoming available in the market. Automation levels 2 and 3 have been given for settled just a couple of years ago, and the path to fully autonomous car seemed to have no obstacles. In reality, OEMs started recently realizing that the impact of semi and/or fully robotized cars on drivers as well as on passengers is all but predictable. VI-grade has more than ten years experience with developing turn-key solution driving simulators, and has been working for more than five years on a research project to collect meaningful bio-signals from the driver during simulator sessions, in collaboration with the BACPIC of the Catholic University of Sacred Heart and the DPIA of the University of Udine. Recently, a collaboration with the Human Inspired Technology Research Center of University of Padova allowed the extension of the assessment at physio-emotional level.

Driving simulators are nowadays a widely used tool in the automotive industry. In particular, the need for safe and repeatable conditions in automated driving testing is now defining a new challenge: to extend the use of the tool to nonprofessional drivers. Quality of the motion control strategies in generating both realistic and feasible inputs to the driver is therefore, more than ever, a crucial aspect. The motion strategies are implemented in the so-called motion cueing algorithms (MCAs). A recently proposed effective approach to MCA is based on model predictive control (MPC), as it is well suited to solve constrained optimal control problems and to take advantage of models of the human sensing system. However, the predictive aspect of the algorithm has not been exploited yet, due to the hard real-time requirement when using long prediction windows. In this paper, a real-time implementation of an MPC-based MCA with predictive feature is presented, endowed with an on-line switching policy to a nonpredictive algorithm when the expected driver behavior is considered unreliable. The motion action based on the actual driver behavior and the expected one are considered in the same procedure, thus fully exploiting the availability of a perceptive model. An optimal tuning procedure is also proposed, based on a multiobjective optimization, where both performance improvement due to the prediction exploitation, and robustness to varying driver behaviour are considered. Finally, a characterization of the driver skill level is proposed and validated in an experimental environment for the specific case of the vertical DOF.

In Nonlinear Model Predictive Control(NMPC), an optimal control problem (OCP) is solved repeatedly at every sampling instant. To satisfy the real-time restriction, modern methods tend to convert the OCP into structured Nonlinear Programming problems (NLP), which are approximately solved on-line. Real-Time Iteration is one of the promising NMPC algorithms that solves the NLP using a single Sequential Quadratic Programming (SQP) iteration. To solve Quadratic Programming (QP) problems efficiently, recently proposed partial condensing techniques reformulate large and sparse QP problems into smaller but still sparse ones. In this paper, we propose two tailored partial condensing algorithms by combining partial sensitivity updating schemes, that are recently proposed to update part of the sensitivities of dynamics. We show that such a combination can significantly reduce the computational time for partial condensing and for the full RTI step.