Dirk Schulz

Efficient multi-resolution plane segmentation of 3d point clouds

We present an efficient multi-resolution approach to segment a 3D point cloud into planar components. In order to gain efficiency, we process large point clouds iteratively from coarse to fine 3D resolutions: At each resolution, we rapidly extract surface normals to describe surface elements (surfels). We group surfels that cannot be associated with planes from coarser resolutions into co-planar clusters with the Hough transform. We then extract connected components on these clusters and determine a best plane fit through RANSAC. Finally, we merge plane segments and refine the segmentation on the finest resolution. In experiments, we demonstrate the efficiency and quality of our method and compare it to other state-of-the-art approaches.

A Component-Based Approach to Visual Person Tracking from a Mobile Platform

In this article, we present a component-based visual tracker for mobile platforms with an application to person tracking. The core of the technique is a component-based descriptor that captures the structure and appearance of a target in a flexible way. This descriptor can be learned quickly from a single training image and is easily adaptable to different objects. It is especially well suited to represent humans since they usually do not have a uniform appearance but, due to clothing, consist of different parts with different appearance. We show how this component-based descriptor can be integrated into a visual tracker based on the well known Condensation algorithm. Several person tracking experiments carried out with a mobile robot in different laboratory environments show that the system is able to follow people autonomously and to distinguish individuals. We furthermore illustrate the advantage of our approach compared to other tracking methods.

SURE: surface entropy for distinctive 3d features

In this paper, we present SURE features --- a novel combination of interest point detector and descriptor for 3D point clouds and depth images. We propose an entropy-based interest operator that selects distinctive points on surfaces. It measures the variation in surface orientation from surface normals in the local vicinity of a point. We complement our approach by the design of a view-pose-invariant descriptor that captures local surface curvature properties, and we propose optional means to incorporate colorful texture information seamlessly. In experiments, we compare our approach to a state-of-the-art feature detector in depth images (NARF) and demonstrate similar repeatability of our detector. Our novel pair of detector and descriptor achieves superior results for matching interest points between images and also requires lower computation time.

Real time interaction with mobile robots using hand gestures

We developed a robust real time hand gesture based interaction system to effectively communicate with a mobile robot which can operate in an outdoor environment. The system enables the user to operate a mobile robot using hand gesture based commands. In particular the system offers direct on site interaction providing better perception of environment to the user. To overcome the illumination challenges in outdoors, the system operates on depth images. Processed depth images are given as input to a convolutional neural network which is trained to detect static hand gestures. The system is evaluated in real world experiments on a mobile robot to show the operational efficiency in outdoor environment.

Fast visual people tracking using a feature-based people detector

This work presents a tracking system that is tailored towards the needs of small mobile robots. It does not rely on person motion estimation in image space because of unpredictable camera movement while the robot traverses non-flat ground. Additionally it is fast enough to process 5 to 10 frames per second which allows interactive applications. The robustness of the tracking system is improved using fused visible spectrum and thermal images. This reduces false positives and enables the system to work in extreme light conditions. Experiments are conducted on both indoor and outdoor data. Outdoor data was recorded at different temperatures and with a moving robot.

Spatially constrained coordinated navigation for a multi-robot system

Abstract In this paper we present a method for navigating a multi-robot system through an environment while additionally maintaining a predefined set of constraints. Examples for constraints are the requirement to maintain a direct line-of-sight between robots or to ensure that the multi-robot system maintains communication. Our approach is based on graph structures that model movements and constraints separately, in order to cover different kind of robots and a large class of possible constraints. Additionally, the separation of movement and constraint graph allows us to use known graph algorithms like Steiner tree heuristics or the multi-point relay algorithm to solve the problem of finding a target configuration for the robots. To connect the movements of the robots with the given constraints, we introduce separated connection graphs which allow assembling valid navigation plans fast. This paper presents some theoretical insight into the problem of coordinated navigation for multi-robot systems with spatial constraints as well as a practical solution. Experiments in simulation and with real robots show the feasibility of the approach.

Design and comparative evaluation of an iterative contact point estimation method for static stability estimation of mobile actively reconfigurable robots

Due to the advancements of robotic systems, they are able to be employed in more unstructured outdoor environments. In such environments the robot-terrain interaction becomes a highly non-linear function. Several methods were proposed to estimate the robot-terrain interaction: machine learning methods, iterative geometric methods, quasi-static and fully dynamic physics simulations. However, to the best of our knowledge there has been no systematic evaluation comparing those methods.In this paper, we present a newly developed iterative contact point estimation method for static stability estimation of actively reconfigurable robots. This new method is systematically compared to a physics simulation in a comprehensive evaluation. Both interaction models determine the contact points between robot and terrain and facilitate a subsequent static stability prediction. Hence, they can be used in our state space global planner for rough terrain to evaluate the robots pose and stability. The analysis also compares deterministic versions of both methods to stochastic versions which account for uncertainty in the robot configuration and the terrain model. The results of this analysis show that the new iterative method is a valid and fast approximate method. It is significantly faster compared to a physics simulation while providing good results in realistic robotic scenarios. We present a new robot pose prediction method for static stability estimation.The method approximates the terrain by least-squares planes to reduce the runtime.A stochastic version accounts for noise in the robot state and the terrain model.We systematically compared it with a physics simulation in many distinct scenarios.The new method is significantly faster and competitive in realistic situations.

Hierarchical rough terrain motion planning using an optimal sampling-based method

Mobile robots with reconfigurable chassis are able to traverse unstructured outdoor environments with boulders or rubble, and overcome challenging structures in urban environments, like stairs or steps. Autonomously traversing rough terrain and such obstacles while ensuring the safety of the robot is a challenging task in mobile robotics. In this paper we introduce a two-phase motion planning algorithm for actively reconfigurable tracked robots. We first use the completeness of a graph search on a regular grid to quickly find an initial path in a low dimensional space, considering only the platforms operating limits instead of the complete state. We then take this initial path to focus the RRT* search in the continuous high-dimensional state space including the actuators of the robot. We do not rely on a detailed structure/terrain classification or use any predefined motion sequences. Hence, our planner can be applied to urban structures, like stairs, as well as rough unstructured environments. Simulation results prove our method to be effective in solving planning queries in such environments.

Motion planning for actively reconfigurable mobile robots in search and rescue scenarios

In disaster scenarios, mobile robots can be employed in hazardous environments where it is too dangerous for human rescuers. Robotic systems can assist rescue personnel as they can be used to explore those inaccessible areas and to assess the situation. Tracked platforms with actuators have been proven to be well suited for such deployments because they are agile enough to overcome quite challenging terrain. A very demanding task for operators is the navigation of the robotic system in complex disaster environments. Hence, an important capability of future systems for search and rescue missions is autonomous navigation in disaster scenarios. In this paper we introduce a two-phase motion planning algorithm for tracked robots with actively controlled actuators to find a fast and stable path to a user specified goal. In the first phase, we generate an initial path considering the platforms operating limits and the terrain roughness. In the second phase, we limit the search space to the area around the initial path and refine the preliminary solution accounting for the complete robot state including actuators and the robots stability and traction. A main distinction of our method is that it does not rely on a previous classification of the terrain, thus, can be applied to a variety of environments. We present experiments evaluating our algorithm in simulation and in two real-world scenarios to demonstrate the validity and feasibility of our approach.

Offroad Navigation Using Adaptable Motion Patterns

We present a navigation system which is able to steer an electronically controlled ground vehicle to given destinations considering all obstacles in its vicinity. The approach is designed for vehicles without a velocity controlled drive-train, making it especially useful for typical remote-controlled vehicles. The vehicle is controlled by sets of commands, each set representing a specific maneuver. These sets are combined in a tree-building procedure to form trajectories towards the given destination. While the sets of commands are executed the vehicle’s behavior is measured to refine the prediction used for path generation. This enables the approach to adapt to surface alterations. We tested our system using a 400xa0kg EOD robot in an outdoor environment.