Institut für Systemdynamik - ISD
Multi-object tracking filters require a birth density to detect new objects from measurement data. If the initial positions of new objects are unknown, it may be useful to choose an adaptive birth density. In this paper, a circular birth density is proposed, which is placed like a band around the surveillance area. This allows for 360° coverage. The birth density is described in polar coordinates and considers all point-symmetric quantities such as radius, radial velocity and tangential velocity of objects entering the surveillance area. Since it is assumed that these quantities are unknown and may vary between different targets, detected trajectories, and in particular their initial states, are used to estimate the distribution of initial states. The adapted birth density is approximated as a Gaussian mixture, so that it can be used for filters operating on Cartesian coordinates.
In many industrial applications a workpiece is continuously fed through a heating zone in order to reach a desired temperature to obtain specific material properties. Many examples of such distributed parameter systems exist in heavy industry and also in furniture production such processes can be found. In this paper, a real-time capable model for a heating process with application to industrial furniture production is modeled. As the model is intended to be used in a Model Predictive Control (MPC) application, the main focus is to achieve minimum computational runtime while maintaining a sufficient amount of accuracy. Thus, the governing Partial Differential Equation (PDE) is discretized using finite differences on a grid, specifically tailored to this application. The grid is optimized to yield acceptable accuracy with a minimum number of grid nodes such that a relatively low order model is obtained. Subsequently, an explicit Runge-Kutta ODE (Ordinary Differential Equation) solver of fourth order is compared to the Crank-Nicolson integration scheme presented in Weiss et al. (2022) in terms of runtime and accuracy. Finally, the unknown thermal parameters of the process are estimated using real-world measurement data that was obtained from an experimental setup. The final model yields acceptable accuracy while at the same time shows promising computation time, which enables its use in an MPC controller.
This paper describes an early lumping approach for generating a mathematical model of the heating process of a moving dual-layer substrate. The heat is supplied by convection and nonlinearly distributed over the whole considered spatial extend of the substrate. Using CFD simulations as a reference, two different modelling approaches have been investigated in order to achieve the most suitable model type. It is shown that due to the possibility of using the transition matrix for time discretization, an equivalent circuit model achieves superior results when compared to the Crank-Nicolson method. In order to maintain a constant sampling time for the in-visioned-control strategies, the effect of variable speed is transformed into a system description, where the state vector has constant length but a variable number of non-zero entries. The handling of the variable transport speed during the heating process is considered as the main contribution of this work. The result is a model, suitable for being used in future control strategies.
Modeling a suitable birth density is a challenge when using Bernoulli filters such as the Labeled Multi-Bernoulli (LMB) filter. The birth density of newborn targets is unknown in most applications, but must be given as a prior to the filter. Usually the birth density stays unchanged or is designed based on the measurements from previous time steps.
In this paper, we assume that the true initial state of new objects is normally distributed. The expected value and covariance of the underlying density are unknown parameters. Using the estimated multi-object state of the LMB and the Rauch-Tung-Striebel (RTS) recursion, these parameters are recursively estimated and adapted after a target is detected.
The main contribution of this paper is an algorithm to estimate the parameters of the birth density and its integration into the LMB framework. Monte Carlo simulations are used to evaluate the detection driven adaptive birth density in two scenarios. The approach can also be applied to filters that are able to estimate trajectories.
This paper presents a modeling approach of an industrial heating process where a stripe-shaped workpiece is heated up to a specific temperature by applying hot air through a nozzle. The workpiece is moving through the heating zone and is considered to be of infinite length. The speed of the substrate is varying over time. The derived model is supposed to be computationally cheap to enable its use in a model-based control setting. We start by formulating the governing PDE and the corresponding boundary conditions. The PDE is then discretized on a spatial grid using finite differences and two different integration schemes, explicit and implicit, are derived. The two models are evaluated in terms of computational effort and accuracy. It turns out that the implicit approach is favorable for the regarded process. We optimize the grid of the model to achieve a low number of grid nodes while maintaining a sufficient amount of accuracy. Finally, the thermodynamical parameters are optimized in order to fit the model's output to real-world data that was obtained by experiments.
A constructive method for the design of nonlinear observers is discussed. To formulate conditions for the construction of the observer gains, stability results for nonlinear singularly perturbed systems are utilised. The nonlinear observer is designed directly in the given coordinates, where the error dynamics between the plant and the observer becomes singularly perturbed by a high-gain part of the observer injection, and the information of the slow manifold is exploited to construct the observer gains of the reduced-order dynamics. This is in contrast to typical high-gain observer approaches, where the observer gains are chosen such that the nonlinearities are dominated by a linear system. It will be demonstrated that the considered approach is particularly suited for self-sensing electromechanical systems. Two variants of the proposed observer design are illustrated for a nonlinear electromagnetic actuator, where the mechanical quantities, i.e. the position and the velocity, are not measured
The trajectory tracking problem for a fully-actuated real-scaled surface vessel is addressed in this paper by designing a backstepping controller with a multivariable integral action, considering the thruster allocation problem. The performance and robustness of this controller are evaluated in simulation, taking into account environmental disturbance forces and modeling mismatch, using a docking maneuver as a reference trajectory. Furthermore, a comparison between the backstepping controller and a nonlinear position PID-Control with flatness based-feedforward is also analyzed.
Comparison and Identifiability Analysis of Friction Models for the Dither Motion of a Solenoid
(2018)
In this paper, the mechanical subsystem of a proportional solenoid excited by a dither signal is considered. The objective is to find a suitable friction model that reflects the characteristic mechanical properties of the dynamic system. Several different friction models from the literature are compared. The friction models are evaluated with respect to their accuracy as well as their practical identifiability, the latter being quantified based on the Fisher information matrix.
In this paper, a systematic comparison of three different advanced control strategies for automated docking of a vessel is presented. The controllers are automatically tuned offline by applying an optimization process using simulations of the whole system including trajectory planner and state and disturbance observer. Then investigations are conducted subject to performance and robustness using Monte Carlos simulation with varying model parameters and disturbances. The control strategies have also been tested in full scale experiments using the solar research vessel Solgenia. The investigated control strategies all have demonstrated very good performance in both, simulation and real world experiments. Videos are available under https://www.htwg-konstanz.de/forschung-und-transfer/institute-und-labore/isd/regelungstechnik/videos/
This paper presents a systematic comparison of different advanced approaches for motion prediction of vessels for docking scenarios. Therefore, a conventional nonlinear gray-box-model, its extension to a hybrid model using an additional regression neural network (RNN) and a black-box-model only based on a RNN are compared. The optimal hyperparameters are found by grid search. The training and validation data for the different models is collected in full-scale experiments using the solar research vessel Solgenia. The performances of the different prediction models are compared in full-scale scenarios. %To use the investigated approaches for controller design, a general optimal control problem containing the advanced models is described. These can improve advanced control strategies e.g., nonlinear model predictive control (NMPC) or reinforcement learning (RL). This paper explores the question of what the advantages and disadvantages of the different presented prediction approaches are and how they can be used to improve the docking behavior of a vessel.
In this paper, the problem of controlling the dissolved oxygen level (DO) during an aerobic fermentation is considered. The proposed approach deals with three major difficulties in respect to the nonlinear dynamics of the DO, the poor accuracy of the empirical models for the oxygen consumption rate and the fact that only sampled measurements are available on-line. A nonlinear integral high-gain control law including a continuous-discrete time observer is designed to keep the DO in the neighborhood of a set point value without any knowledge on the dissolved oxygen consumption rate. The local stability of the control algorithm is proved using Lyapunov tools. The performance of the control scheme is first analyzed in simulation and then experimentally evaluated during a successfull fermentation of the bacteria over a period of three days. Pseudomonas putida mt-2
Docking Control of a Fully-Actuated Autonomous Vessel using Model Predictive Path Integral Control
(2022)
This paper presents the docking control of an autonomous vessel using the nonlinear Model Predictive Path Integral (MPPI) approach. This algorithm is based on a path integral over stochastic trajectories and can be parallelized easily. The controller parameters are tuned offline using knowledge of the system and simulations, including nonlinear state and disturbance observer. The cost function implicitly contains information regarding the surrounding of the docking position. This approach allows continuous optimization of the trajectory with respect to the system state, disturbance state and actuator dynamics. The control strategy has been tested in full-scale experiments using the solar research vessel Solgenia. The investigated MPPI controller has demonstrated excellent performance in both, simulation and real-world experiments. This paper addresses the question of how the MPPI algorithm can be applied to dock a fully-actuated vessel and what benefits its application achieves.
This paper compares novel methods to efficiently include input constraints using the nonlinear Model Predictive Path Integral (MPPI) approach. The MPPI algorithm solves stochastic optimal control problems and is based on sampled trajectories. MPPI results from the physical path integral framework. Sample-based algorithms are characterized by the fact that they can be computed in parallel and offer the possibility to handle discontinuous dynamics and cost functions. However, using standard MPPI the input costs in the Lagrange term have to be chosen quadratic. This fact is unfavorable for various real applications. Further, in standard nonlinear model predictive control (NMPC) approaches hard box constraints on the control input trajectory can be treated directly. In this contribution, novel architectures based on integrator action are compared. The investigated input constraint MPPI controllers were tested on an autonomous self-balancing vehicle. Therefore both, simulation and real-world experiments are presented. This paper addresses the question of how the MPPI algorithm can be further developed to consider input box constraints. Videos of the self-balancing vehicle are available at: https: https://tinyurl.com/mvn8j7vf
Experimental Validation of Ellipsoidal Techniques for State Estimation in Marine Applications
(2022)
A reliable quantification of the worst-case influence of model uncertainty and external disturbances is crucial for the localization of vessels in marine applications. This is especially true if uncertain GPS-based position measurements are used to update predicted vessel locations that are obtained from the evaluation of a ship’s state equation. To reflect real-life working conditions, these state equations need to account for uncertainty in the system model, such as imperfect actuation and external disturbances due to effects such as wind and currents. As an application scenario, the GPS-based localization of autonomous DDboat robots is considered in this paper. Using experimental data, the efficiency of an ellipsoidal approach, which exploits a bounded-error representation of disturbances and uncertainties, is demonstrated.
The random matrix approach is a robust algorithm to filter the mean and covariance matrix of noisy observations of a dynamic object. Afterward, virtual measurement models can be used to find iteratively the extent parameters of an object that would cause the same statistical moments within their measurements. In previous work, this was limited to elliptical targets and only contour measurements.In this paper, we introduce the parallel use of an elliptical, triangular and rectangular-shaped virtual measurement model and a shape classification that selects the model that fits best to the measurements. The measurement likelihood is modeled either via ray tracing, a uniformly or normally spatial distribution over the object’s extent or as a combination of those.The results show that the extent estimation works precisely and that the classification accuracy highly depends on the measurement noise.
Random matrices are used to filter the center of gravity (CoG) and the covariance matrix of measurements. However, these quantities do not always correspond directly to the position and the extent of the object, e.g. when a lidar sensor is used.In this paper, we propose a Gaussian processes regression model (GPRM) to predict the position and extension of the object from the filtered CoG and covariance matrix of the measurements. Training data for the GPRM are generated by a sampling method and a virtual measurement model (VMM). The VMM is a function that generates artificial measurements using ray tracing and allows us to obtain the CoG and covariance matrix that any object would cause. This enables the GPRM to be trained without real data but still be applied to real data due to the precise modeling in the VMM. The results show an accurate extension estimation as long as the reality behaves like the modeling and e.g. lidar measurements only occur on the side facing the sensor.
Extended Target Tracking With a Lidar Sensor Using Random Matrices and a Virtual Measurement Model
(2022)
Random matrices are widely used to estimate the extent of an elliptically contoured object. Usually, it is assumed that the measurements follow a normal distribution, with its standard deviation being proportional to the object’s extent. However, the random matrix approach can filter the center of gravity and the covariance matrix of measurements independently of the measurement model. This work considers the whole chain from data acquisition to the linear Kalman Filter with extension estimation as a reference plant. The input is the (unknown) ground truth (position and extent). The output is the filtered center of gravity and the filtered covariance matrix of the measurement distribution. A virtual measurement model emulates the behavior of the reference plant. The input of the virtual measurement model is adapted using the proposed algorithm until the output parameters of the virtual measurement model match the result of the reference plant. After the adaptation, the input to the virtual measurement model is considered an estimation for position and extent. The main contribution of this paper is the reference model concept and an adaptation algorithm to optimize the input of the virtual measurement model.
In this paper, approximating the shape of a sailing boat using elliptic cones is investigated. Measurements are assumed to be gathered from the target's surface recorded by 3D scanning devices such as multilayer LiDAR sensors. Therefore, different models for estimating the sailing boat's extent are presented and evaluated in simulated and real-world scenarios. In particular, the measurement source association problem is addressed in the models. Simulated investigations are conducted with a static and a moving elliptic cone. The real-world scenario was recorded with a Velodyne Alpha Prime (VLP-128) mounted on a ferry of Lake Constance. Final results of this paper constitute the extent estimation of a single sailing boat using LiDAR data applying various measurement models.
In this paper, a novel feature-based sampling strategy for nonlinear Model Predictive Path Integral (MPPI) control is presented. Using the MPPI approach, the optimal feedback control is calculated by solving a stochastic optimal control (OCP) problem online by evaluating the weighted inference of sampled stochastic trajectories. While the MPPI algorithm can be excellently parallelized, the closed-loop performance strongly depends on the information quality of the sampled trajectories. To draw samples, a proposal density is used. The solver’s and thus, the controller’s performance is of high quality if the sampled trajectories drawn from this proposal density are located in low-cost regions of state-space. In classical MPPI control, the explored state-space is strongly constrained by assumptions that refer to the control value’s covariance matrix, which are necessary for transforming the stochastic Hamilton–Jacobi–Bellman (HJB) equation into a linear second-order partial differential equation. To achieve excellent performance even with discontinuous cost functions, in this novel approach, knowledge-based features are introduced to constitute the proposal density and thus the low-cost region of state-space for exploration. This paper addresses the question of how the performance of the MPPI algorithm can be improved using a feature-based mixture of base densities. Furthermore, the developed algorithm is applied to an autonomous vessel that follows a track and concurrently avoids collisions using an emergency braking feature. Therefore, the presented feature-based MPPI algorithm is applied and analyzed in both simulation and full-scale experiments.
Feature-Based Proposal Density Optimization for Nonlinear Model Predictive Path Integral Control
(2022)
This paper presents a novel feature-based sampling strategy for nonlinear Model Predictive Path Integral (MPPI) control. In MPPI control, the optimal control is calculated by solving a stochastic optimal control problem online using the weighted inference of stochastic trajectories. While the algorithm can be excellently parallelized the closed- loop performance is dependent on the information quality of the drawn samples. Because these samples are drawn using a proposal density, its quality is crucial for the solver and thus the controller performance. In classical MPPI control, the explored state-space is strongly constrained by assumptions that refer to the control value variance, which are necessary for transforming the Hamilton-Jacobi-Bellman (HJB) equation into a linear second-order partial differential equation. To achieve excellent performance even with discontinuous cost-functions, in this novel approach, knowledge-based features are used to determine the proposal density and thus, the region of state- space for exploration. This paper addresses the question of how the performance of the MPPI algorithm can be improved using a feature-based mixture of base densities. Further, the developed algorithm is applied on an autonomous vessel that follows a track and concurrently avoids collisions using an emergency braking feature.