Stefan Kohlbrecher

A flexible and scalable SLAM system with full 3D motion estimation

For many applications in Urban Search and Rescue (USAR) scenarios robots need to learn a map of unknown environments. We present a system for fast online learning of occupancy grid maps requiring low computational resources. It combines a robust scan matching approach using a LIDAR system with a 3D attitude estimation system based on inertial sensing. By using a fast approximation of map gradients and a multi-resolution grid, reliable localization and mapping capabilities in a variety of challenging environments are realized. Multiple datasets showing the applicability in an embedded hand-held mapping system are provided. We show that the system is sufficiently accurate as to not require explicit loop closing techniques in the considered scenarios. The software is available as an open source package for ROS.

Comprehensive simulation of quadrotor UAVs using ROS and gazebo

Quadrotor UAVs have successfully been used both in research and for commercial applications in recent years and there has been significant progress in the design of robust control software and hardware. Nevertheless, testing of prototype UAV systems still means risk of damage due to failures. Motivated by this, a system for the comprehensive simulation of quadrotor UAVs is presented in this paper. Unlike existing solutions, the presented system is integrated with ROS and the Gazebo simulator. This comprehensive approach allows simultaneous simulation of diverse aspects such as flight dynamics, onboard sensors like IMUs, external imaging sensors and complex environments. The dynamics model of the quadrotor has been parameterized using wind tunnel tests and validated by a comparison of simulated and real flight data. The applicability for simulation of complex UAV systems is demonstrated using LIDAR-based and visual SLAM approaches available as open source software.

Vision based victim detection from unmanned aerial vehicles

Finding injured humans is one of the primary goals of any search and rescue operation. The aim of this paper is to address the task of automatically finding people lying on the ground in images taken from the on-board camera of an unmanned aerial vehicle (UAV). In this paper we evaluate various state-of-the-art visual people detection methods in the context of vision based victim detection from an UAV. The top performing approaches in this comparison are those that rely on flexible part-based representations and discriminatively trained part detectors. We discuss their strengths and weaknesses and demonstrate that by combining multiple models we can increase the reliability of the system. We also demonstrate that the detection performance can be substantially improved by integrating the height and pitch information provided by on-board sensors. Jointly these improvements allow us to significantly boost the detection performance over the current de-facto standard, which provides a substantial step towards making autonomous victim detection for UAVs practical.

Human-robot Teaming for Rescue Missions: Team ViGIR's Approach to the 2013 DARPA Robotics Challenge Trials

Team ViGIR entered the 2013 DARPA Robotics Challenge DRC with a focus on developing software to enable an operator to guide a humanoid robot through the series of challenge tasks emulating disaster response scenarios. The overarching philosophy was to make our operators full team members and not just mere supervisors. We designed our operator control station OCS to allow multiple operators to request and share information as needed to maintain situational awareness under bandwidth constraints, while directing the robot to perform tasks with most planning and control taking place onboard the robot. Given the limited development time, we leveraged a number of open source libraries in both our onboard software and our OCS design; this included significant use of the robot operating system libraries and toolchain. This paper describes the high level approach, including the OCS design and major onboard components, and it presents our DRC Trials results. The paper concludes with a number of lessons learned that are being applied to the final phase of the competition and are useful for related projects as well.

Hector Open Source Modules for Autonomous Mapping and Navigation with Rescue Robots

Key abilities for robots deployed in urban search and rescue tasks include autonomous exploration of disaster sites and recognition of victims and other objects of interest. In this paper, we present related open source software modules for the development of such complex capabilities which include hector_slam for self-localization and mapping in a degraded urban environment. All modules have been successfully applied and tested originally in the RoboCup Rescue competition. Up to now they have already been re-used and adopted by numerous international research groups for a wide variety of tasks. Recently, they have also become part of the basis of a broader initiative for key open source software modules for urban search and rescue robots.

Supervised footstep planning for humanoid robots in rough terrain tasks using a black box walking controller

In recent years, the numbers of life-size humanoids as well as their mobility capabilities have steadily grown. Stable walking motion and control for humanoid robots are already well investigated research topics. This raises the question how navigation problems in complex and unstructured environments can be solved utilizing a given black box walking controller with proper perception and modeling of the environment provided. In this paper we present a complete system for supervised footstep planning including perception, world modeling, 3D planner and operator interface to enable a humanoid robot to perform sequences of steps to traverse uneven terrain. A proper height map and surface normal estimation are directly obtained from point cloud data. A search-based planning approach (ARA*) is extended to sequences of footsteps in full 3D space (6 DoF). The planner utilizes a black box walking controller without knowledge of its implementation details. Results are presented for an Atlas humanoid robot during participation of Team ViGIR in the 2013 DARPA Robotics Challenge Trials.

Template-based manipulation in unstructured environments for supervised semi-autonomous humanoid robots

Humanoid robotic manipulation in unstructured environments is a challenging problem. Limited perception, communications and environmental constraints present challenges that prevent fully autonomous or purely teleoperated robots from reliably interacting with their environment. In order to achieve higher reliability in manipulation we present an approach involving remote human supervision. Strengths from both human operator and humanoid robot are leveraged through a user interface that allows the operator to perceive the remote environment through an aggregated worldmodel based on onboard sensing, while the robot can efficiently receive perceptual and semantic information from the operator. A template based manipulation approach has been successfully applied to the Atlas humanoid robot; we show real world footage of the results obtained in the DARPA Robotics Challenge Trials 2013.

Human-robot collaborative high-level control with application to rescue robotics

Motivated by the DARPA Robotics Challenge (DRC), the application of operator assisted (semi-)autonomous robots with highly complex locomotion and manipulation abilities is considered for solving complex tasks in potentially unknown and unstructured environments. Because of the limited a priori knowledge about the state of the environment and tasks needed to achieve a complex mission, a sufficiently complete a priori design of high level robot behaviors is not possible. Most of the situational knowledge required for such behavior design is gathered only during runtime and needs to be interpreted by a human operator. However, current behavior control approaches only allow for very limited adaptation at runtime and no flexible operator interaction. In this paper an approach for definition and execution of complex robot behaviors based on hierarchical state machines is presented, allowing to flexibly change the structure of behaviors on the fly during runtime through assistance of a remote operator. The efficiency of the proposed approach is demonstrated and evaluated not only in an example scenario, but also by application in two robot competitions.

Achieving versatile manipulation tasks with unknown objects by supervised humanoid robots based on object templates

The investigations of this paper are motivated by the scenario of a supervised semi-autonomous humanoid robot entering a mainly unknown, potentially degraded human environment to perform highly diverse disaster recovery tasks. For this purpose, the robot must be enabled to use any object it can find in the environment as tool for achieving its current manipulation task. This requires the use of potential unknown objects as well as known objects for new purposes (e.g. using a drill as a hammer). A recently proposed object template manipulation approach is extended to provide a semi-autonomous humanoid robot assisted by a remote human supervisor with the versatility needed to utilize objects in the described manner applying affordances [1] from other previously known objects. For an Atlas humanoid robot it is demonstrated how using a small set of such object templates with well defined affordances can be used to solve manipulation tasks using new unknown objects.

Overview of team ViGIR's approach to the Virtual Robotics Challenge

With the DARPA Robotics Challenge (DRC), a call to an ambitious multi-part competition was sent out to the robotics community. In this paper, we briefly summarize the approach for addressing the Virtual Robotics Challenge (VRC) where software for control and supervision of a capable humanoid robot must be developed. Team ViGIR, comprising members from the US and Germany, leveraged previous robotics competition experience and a variety of open source tools, to achieve sixth place in the VRC out of 126 registrants, thereby advancing to the next round of the DRC and obtaining an Atlas robot.