Wendelin Feiten

An Atlas framework for scalable mapping

This paper describes Atlas, a hybrid metrical/topological approach to SLAM that achieves efficient mapping of large-scale environments. The representation is a graph of coordinate frames, with each vertex in the graph representing a local frame, and each edge representing the transformation between adjacent frames. In each frame, we build a map that captures the local environment and the current robot pose along with the uncertainties of each. Each maps uncertainties are modeled with respect to its own frame. Probabilities of entities with respect to arbitrary frames are generated by following a path formed by the edges between adjacent frames, computed via Dijkstras shortest path algorithm. Loop closing is achieved via an efficient map matching algorithm. We demonstrate the technique running in real-time in a large indoor structured environment (2.2 km path length) with multiple nested loops using laser or ultrasonic ranging sensors.

Field test of a navigation system: autonomous cleaning in supermarkets

Siemens has developed a prototype of an autonomous navigation system for mobile service robots. Its suitability for tough everyday operation has been successfully demonstrated since August 1996 with cleaning machines in chain store supermarkets of Albert Heijn BV in the Netherlands. Here, HEFTER CLEANTECH (HCT) cleaning robots are being used in cooperation with the cleaning specialist RTB.

Robust obstacle avoidance in unknown and cramped environments

This paper presents a novel local obstacle avoidance module. The main idea of the steer angle field approach is to discard all steer angles that, given the nonholonomic kinematics of the robot, would lead to collisions within a certain hit distance. The approach has been implemented and tested over a period of one year on our mobile robot. Results have shown that the robot is able to operate robustly in unknown, unprepared and cramped in-door environments for hours.<<ETX>>

Steer angle fields: An approach to robust manoeuvring in cluttered, unknown environments

Abstract In this paper a simple, robust and efficient method for local obstacle avoidance is described. It takes into account both the geometric and the kinematic properties of the robot in order to calculate allowed and forbidden steer angles. The algorithm has been implemented and extensively tested on our mobile robot. Results have shown that the robot is able to operate robustly in unknown and uprepared in-door environments for extended periods.

Sonar sensing for low-cost indoor mobility

Abstract Low-cost autonomous mobility poses an important challenge on the way to new robot applications. Ultrasonic sensor systems are both robust and low-cost. We therefore have investigated the usefulness of several types of ultrasonic sensor principles for obstacle avoidance and localisation. Our experiments show that by tailoring the algorithms to the specific sensing task at hand, autonomous mobility can to a wide extent be achieved with a simple monaural system. However, some drawbacks like limited speed and resolution in map building remain. These drawbacks can be overcome by more advanced ultrasonic sensor systems, where two or three receivers instead of one are used in order to get more data from one single measurement.

A Probabilistic Relational Model for Characterizing Situations in Dynamic Multi-Agent Systems

Artificial systems with a high degree of autonomy require reliable semantic information about the context they operate in. State interpretation, however, is a difficult task. Interpretations may depend on a history of states and there may be more than one valid interpretation. We propose a model for spatio-temporal situations using hidden Markov models based on relational state descriptions, which are extracted from the estimated state of an underlying dynamic system. Our model covers concurrent situations, scenarios with multiple agents, and situations of varying durations. To evaluate the practical usefulness of our model, we apply it to the concrete task of online traffic analysis.

Reactive motion control for human-robot tactile interaction

In the field of service robotics, robots serve and assist human beings. It is natural for humans to directly interact with the robot via tactile interfaces. This paper introduces several kinds of tactile interactions between a user and the robot as well as interactions of the robot with the environment. All interactions are implemented in a single paradigm: Forces measured from tactile sensors result in motion vectors at the contact points. The motion vectors from different sensors are superimposed and then determine the robots joint velocities. We present results from our experimental setup consisting of an 8 degrees of freedom manipulator arm mounted on a mobile platform. In the illustrated example, a human interacts with the robot using only the tactile interface.

6D Pose Uncertainty in Robotic Perception

Robotic perception is fundamental to important application areas. In the Joint Research Project DESIRE, we develop a robotic perception system with the aim of perceiving and modeling an unprepared kitchen scenario with many objects. It relies on the fusion of information from weak features from heterogenous sensors in order to classify and localize objects. This requires the representation of wide spread probability distributions of the 6D pose.

Fast parametric viewpoint estimation for active object detection

Most current solutions to active perception planning struggle with complex state representations or fast and efficient sensor parameter selection strategies. The goal is to find new viewpoints or optimize sensor parameters for further measurements in order to classify an object and precisely locate its position. This paper presents an exclusively parametric approach for the state estimation and decision making process to achieve very low computational complexity and short calculation times. The proposed approach assumes a realistic, high dimensional and continuous state space for the representation of objects expressing their rotation, translation and class. Its probability distribution is described by multivariate mixtures of Gaussians which allow the representation of arbitrary object hypotheses. In a statistical framework Bayesian state estimation updates the current state probability distribution based on a scene observation which depends on the sensor parameters. These are selected in a decision process which aims on reducing the uncertainty in the state distribution. Approximations of information theoretic measurements are used as evaluation criteria.

Efficient trajectory optimization using a sparse model

The “timed elastic band” approach optimizes robot trajectories by subsequent modification of an initial trajectory generated by a global planner. The objectives considered in the trajectory optimization include but are not limited to the overall path length, trajectory execution time, separation from obstacles, passing through intermediate way points and compliance with the robots dynamic, kinematic and geometric constraints. “Timed elastic bands” explicitly consider spatial-temporal aspects of the motion in terms of dynamic constraints such as limited robot velocities and accelerations. The trajectory planning operates in real time such that “timed elastic bands” cope with dynamic obstacles and motion constraints. The “timed elastic band problem” is formulated as a scalarized multi-objective optimization problem. Most objectives are local and relate to only a small subset of parameters as they only depend on a few consecutive robot states. This local structure results in a sparse system matrix, which allows the utilization of fast and efficient optimization techniques such as the open-source framework “g2o” for solving “timed elastic band” problems. The “g2o” sparse system solvers have been successfully applied to VSLAM problems. This contribution describes the application and adaptation of the g2o-framework in the context of trajectory modification with the “timed elastic band”. Results from simulations and experiments with a real robot demonstrate that the implementation is robust and computationally efficient.