Joris De Schutter

Kalman filters for non-linear systems: a comparison of performance

The Kalman filter is a well-known recursive state estimator for linear systems. In practice, the algorithm is often used for non-linear systems by linearizing the systems process and measurement models. Different ways of linearizing the models lead to different filters. In some applications, these ‘Kalman filter variants’ seem to perform well, while for other applications they are useless. When choosing a filter for a new application, the literature gives us little to rely on. This paper tries to bridge the gap between the theoretical derivation of a Kalman filter variant and its performance in practice when applied to a non-linear system, by providing an application-independent analysis of the performances of the common Kalman filter variants.  Correlated uncertainties can be dealt with by augmenting the state vector, this is the original formulation of the KF (Kalman 1960). Expressed in this new state vector, the process and measurement models are of the form (1) and (2) with uncorrelated uncertainties. This paper separates performance evaluation of Kalman filters into (i) consistency, and (ii) information content of the estimates; and it separates the filter structure into (i) the process update step, and (ii) the measurement update step. This decomposition provides the insights supporting an objective and systematic evaluation of the appropriateness of a particular Kalman filter variant in a particular application.

Constraint-based Task Specification and Estimation for Sensor-Based Robot Systems in the Presence of Geometric Uncertainty

This paper introduces a systematic constraint-based approach to specify complex tasks of general sensor-based robot systems consisting of rigid links and joints. The approach integrates both instantaneous task specification and estimation of geometric uncertainty in a unified framework. Major components are the use of feature coordinates, defined with respect to object and feature frames, which facilitate the task specification, and the introduction of uncertainty coordinates to model geometric uncertainty. While the focus of the paper is on task specification, an existing velocity- based control scheme is reformulated in terms of these feature and uncertainty coordinates. This control scheme compensates for the effect of time varying uncertainty coordinates. Constraint weighting results in an invariant robot behavior in case of conflicting constraints with heterogeneous units. The approach applies to a large variety of robot systems (mobile robots, multiple robot systems, dynamic human-robot interaction, etc.), various sensor systems, and different robot tasks. Ample simulation and experimental results are presented.

Brief paper: Robust high-order repetitive control: Optimal performance trade-offs

High-order repetitive control has previously been introduced to either improve the robustness for period-time uncertainty or reduce the sensitivity for non-periodic inputs of standard repetitive control schemes. This paper presents a systematic, semidefinite programming based approach to compute high-order repetitive controllers that yield an optimal trade-off between these two performance criteria. The methodology is numerically illustrated through trade-off curves for various controller orders and levels of period-time uncertainty. Moreover, existing high-order repetitive control approaches are shown to correspond to specific points on these curves.

Estimating first-order geometric parameters and monitoring contact transitions during force controlled compliant motion

This paper uses (linearized) Kalman filters to estimate first-order geometric parameters (i.e., orientation of contact normals and location of contact points) that occur in force-controlled compliant motions. The time variance of these parameters is also estimated. In addition, transitions between contact situations can be monitored. The contact between the manipulated object and its environment is general, i.e., multiple contacts can occur at the same time, and both the topology and the geometry of each single contact are arbitrary. The two major theoretical contributions are 1) the integration of the general contact model, developed previously by the authors, into a state-space form suitable for recursive processing; and 2) the use of the reciprocity constraint between ideal contact forces and motion freedoms as the “measurement equation” of the Kalman filter. The theory is illustrated by full 3-D experiments. The approach of this paper allows a breakthrough in the state of the art dominated by the classical, orthogonal contact models of Mason that can only cope with a limited (albeit important) subset of all possible contact situations.

An adaptable system for RGB-D based human body detection and pose estimation

HighlightsDoes not require pre-processing by background subtraction and no initialization poses.Online learned appearance model combining color with depth-based labeling.Works in clutter and with body part occlusions because of underlying kinematic model.RDF training, data generation and cluster-based learning, that enables retraining. Human body detection and pose estimation is useful for a wide variety of applications and environments. Therefore a human body detection and pose estimation system must be adaptable and customizable. This paper presents such a system that extracts skeletons from RGB-D sensor data. The system adapts on-line to difficult unstructured scenes taken from a moving camera (since it does not require background subtraction) and benefits from using both color and depth data. It is customizable by virtue of requiring less training data, having a clearly described training method, and a customizable human kinematic model. Results show successful application to data from a moving camera in cluttered indoor environments. This system is open-source, encouraging reuse, comparison, and future research.

Contact-State Segmentation Using Particle Filters for Programming by Human Demonstration in Compliant-Motion Tasks

This paper presents a contribution to programming by human demonstration, in the context of compliant-motion task specification for sensor-controlled robot systems that physically interact with the environment. One wants to learn about the geometric parameters of the task and segment the total motion executed by the human into subtasks for the robot, that can each be executed with simple compliant-motion task specifications. The motion of the human demonstration tool is sensed with a 3-D camera, and the interaction with the environment is sensed with a force sensor in the human demonstration tool. Both measurements are uncertain, and do not give direct information about the geometric parameters of the contacting surfaces, or about the contact formations (CFs) encountered during the human demonstration. The paper uses a Bayesian sequential Monte Carlo method (also known as a particle filter) to do the simultaneous estimation of the CF (discrete information) and the geometric parameters (continuous information). The simultaneous CF segmentation and the geometric parameter estimation are helped by the availability of a contact state graph of all possible CFs. The presented approach applies to all compliant-motion tasks involving polyhedral objects with a known geometry, where the uncertain geometric parameters are the poses of the objects. This work improves the state of the art by scaling the contact estimation to all possible contacts, by presenting a prediction step based on the topological information of a contact state graph, and by presenting efficient algorithms that allow the estimation to operate in real time. In real-world experiments, it is shown that the approach is able to discriminate in real time between some 250 different CFs in the graph

Integrated Vision/Force Robotic Servoing in the Task Frame Formalism

In this paper we show how to use the task frame to easily model, implement and execute three-dimensional (3D) robotic tasks, which integrate force control and visual servoing, in an uncalibrated workspace. In contrast to most hybrid vision/force research, this work uses eye-in-hand vision and force control. Mounting both sensors on the same end-effector gives rise to new constraints, control issues and advantages, which are discussed in this paper. On the one hand, in this paper we emphasize shared control in which both vision and force simultaneously control a given task frame direction. Our work shows the usefulness and feasibility of a range of tasks which use shared control. Moreover, it offers a framework based on the task frame formalism (TFF) to distinguish between different basic forms of shared control. Each basic form is illustrated by a robotic task with shared control in only one direction. In addition, an extension to classify multi-dimensional shared control tasks is presented. On the other hand, a new classification is presented which distinguishes between four meaningful tool/camera configurations, being parallel or non-parallel endpoint closed-loop and fixed or variable endpoint open-loop. Corresponding control strategies are discussed, resulting in either collocated or non-collocated vision/force control. Several task examples (in 3D space), specified in the TFF, illustrate the use of these four configurations. As shown by the presented experimental results, the tasks at hand benefit from the integrated control approach.

Nonlinear Kalman filtering for force-controlled robot tasks

Introduction.- Literature Survey: Autonomous Compliant Motion.- Literature Survey: Bayesian Probability Theory.- Kalman Filters for Nonlinear Systems.- The Non-Minimal State Kalman Filter.- Contact Modelling.- Geometrical Parameter Estimation and CF Recognition.- Experiment: A Cube-In-Corner Assembly.- Task Planning with Active Sensing.- General Conclusions.

Extending iTaSC to support inequality constraints and non-instantaneous task specification

In [1], we presented our constraint-based programming approach, iTaSC1, that formulates instantaneous sensor-based robot tasks as constraint sets, and subsequently solves a corresponding least-squares problem to obtain control set points, such as desired joint velocities or joint torques. This paper further extends this approach, (i) by explicitly supporting the inclusion of inequality constraints in the task and (ii) by supporting a broader class of objective functions for translating the task constraints into robot motion. These extensions are made while retaining a tractable mathematical problem structure (a convex program). Furthermore, first results on extending the approach to non-instantaneous tasks are presented. As illustrated in the paper, the power of the approach lies (i) at its versatility to specify a wide range of robot behaviors and the ease of making task adjustments, and (ii) at its generic nature, that permits using systematic procedures to derive the underlying control equations.

Polyhedral contact formation identification for autonomous compliant motion: exact nonlinear bayesian filtering

This work presents new experimental results for the estimation of large position and orientation inaccuracies of contacting objects during force-controlled compliant motion. The estimation is based on position, velocity, and force measurements. The authors have described the contact modeling and presented some Kalman filter identification results for small inaccuracies. However, when dealing with larger inaccuracies, the nonlinear estimation problem remained unsolved. This problem has now been solved satisfactorily by applying a new Bayesian estimator. The Bayesian filter is valid for static systems (parameter estimation) with any kind of nonlinear measurement equation, subject to Gaussian measurement uncertainty and for a limited class of dynamic systems. Experimental results of this new filter are given for the estimation of the positions and orientations of contacting objects during the cube-in-corner assembly described in the first reference.