Sebastian Klemm

Unified GPU voxel collision detection for mobile manipulation planning

This paper gives an overview on our framework for efficient collision detection in robotic applications. It unifies different data structures and algorithms that are optimized for Graphics Processing Unit (GPU) architectures. A speed-up in various planning scenarios is achieved by utilizing storage structures that meet specific demands of typical use-cases like mobile platform planning or full body planning. The system is also able to monitor the execution of motion trajectories for intruding dynamic obstacles and triggers a replanning or stops the execution. The presented collision detection is deployed in local dynamic planning with live pointcloud data as well as in global a-priori planning. Three different mobile manipulation scenarios are used to evaluate the performance of our approach.

A multi-resolution 3-D environment model for autonomous planetary exploration

A key skill for autonomous exploration and inspection missions is the ability to find safe and traversable paths within previously unknown environments. We present an approach for mapping typical environments encountered by autonomous planetary exploration robots, a pre-interpreted multi-resolution 3-D environment model generated from point cloud data, and a hybrid planner for basically any kind of mobile robot. Our system builds upon and enhances freely available standard frameworks such as ROS and OMPL. We present results of our system applied to our six-legged walking robot LAURON V, showing the progression from individual 3-D point clouds to a rich environment model queried by an RRT*-based planner to find and adapt a feasible and optimal path.

GPU-based real-time collision detection for motion execution in mobile manipulation planning

In this paper we present a parallel collision checking approach as an essential building block of a reactive online planning framework, that allows to continuously monitor the execution of planned trajectories against dynamic changes in the environment. The software is optimized for massively parallel hardware architectures, namely CUDA GPUs and offers constant runtime regardless of the occupancy density in the environment.

Testing and validating high level components for automated driving: simulation framework for traffic scenarios

Current advances in the research field of autonomous driving demand advanced simulation methods for testing and validation. By combining versatile foci of different simulations, we can provide an increased amount and diversity of realistic traffic scenarios, which are relevant to the development and verification of high level automated driving functions. The focus of the present paper is to propose a concept for realistic simulation scenarios, which is capable of running in different integration levels, from software- to vehicle-in-the-loop. Its application is demonstrated, exposing an experimental vehicle, which is used for autonomous driving development, to a traffic scenario with virtual vehicles on a real road network.

Lane-precise localization of intelligent vehicles using the surrounding object constellation

Advanced Driver Assistance Systems (ADAS) towards autonomous driving require an ego vehicle localization on maps to be able to use the map data for e.g. behavior and trajectory prediction of traffic participants.

A semantic approach to sensor-independent vehicle localization

As intelligent vehicles become more and more capable, they must learn to navigate and localize themselves in a wide variety of environments, including GPS-denied and only crudely mapped areas. We argue that since autonomous vehicles must be able to perceive, and semantically interpret, their immediate environment, they should be able to use abstract semantic information as their sole means of localization. This simplifies the level of detail and precision required from environment maps so that, for example, a rough floor plan of a parking garage will suffice to autonomously navigate it. We propose a concept for semantic localization which only requires a conceptual semantic map of the environment, and can be made to work with any kind of sensor data from which the required semantic information can be extracted. We present a localization algorithm which may be used as a base for semantic navigation, e.g. in context of automated driving, and some initial results of its application in a parking garage scenario.

Autonomous multi-story navigation for valet parking

Electric mobility combined with recent advances in autonomous driving provides a solution to the environmental and traffic challenges of the modern metropolis. In this work we present an innovative system that completely changes valet parking and the process of charging electric vehicles. The introduced system tackles the problem of precise and efficient autonomous navigation for vehicles in gps-denied environments like 3-D multi-story parking garages. In addition a robot is employed to autonomously charge the parked electric vehicles. We give insight into the concept and implementation of such a system, and evaluate it in real parking garages. We extensively tested the system in a real-world application, where a driver leaves the vehicle at the entry of a parking garage and the vehicle then performs the navigation and parking task on its own. Our test vehicle autonomously navigated more than 50 times from the entry of a parking garage to an assigned parking spot on the 6th floor and docked with the charging robot. The navigation system is precise, efficient and capable of running online in real-world scenarios.

Anticipate your surroundings: Predictive collision detection between dynamic obstacles and planned robot trajectories on the GPU

Proactive collision detection enables robots to efficiently execute tasks in shared human-robot-workspaces by avoiding collision-prone situations. Our work connects motion prediction of RGB-D flow algorithms with motion primitive planning via an efficient voxel Swept-Volume-based collision detection. The approach can handle scenarios with varying contents as we use the same techniques to predict single moving objects but also articulated bodies. Our process chain consists of highly parallel GPU algorithms that allow a full 3D representation of planned trajectories and predictions from live pointclouds in high resolution, while still being online capable. We demonstrated our achievements in two scenarios with different motion granularity.

Eye in hand: Towards GPU accelerated online grasp planning based on pointclouds from in-hand sensor

This work proposes the usage of in-hand depth cameras in combination with GPU-based collision detection algorithms to realize robotics grasp planning for unknown objects on the fly. Based on pointcloud data captured during an exploratory hand motion we evaluate and optimize grasps with a hybrid Particle Swarm Optimization process. The approach maximizes the contact surface while examining various grasp poses and allows a precise and model-free manipulation of arbitrary objects. Targeted end-effectors are anthropomatic multi-fingered hands with complex kinematics and geometries.

RoaDS — Robot and dynamics simulation for biologically-inspired multi-legged walking robots

Increasing computational power and efficient physics engines make robotic simulations popular and applied in more and more domains and scenarios. Well established simulators like GAZEBO have become an important part in robotics research. Nevertheless, non of the popular simulators addresses the needs of multi-legged walking robots. In this work, we develop an efficient and precise simulation system for multi-legged robots: RoaDS. The requirements, architecture and distributed design are presented. A series of experiments with the six-legged walking robot LAURON evaluates and confirms its high accuracy and good performance. RoaDS is a great tool to compare designs of bio-inspired robots, analyze walking patterns and improve robot walking skills.