Sense and avoid is one of the actual challenges for unmanned systems. Focus of this paper is the aircraft domain and the goal is to create a system that can act like a pilot see what is around the vehicle, detect obstacles and other imminent dangers, and plan a collision-free trajectory around them. This requires sensors like radar, laser scanners or cameras for environmental perception and the hardware and software for data analysis and automatic control. In principle, there are two ways of obstacle avoidance implementation. The first is adapted from insects and might be more simple since such approaches convert sensor data directly into reactive movement instructions. Successful results have already been presented for aerial vehicles using optical flow, a combination of optical flow and stereo vision, stereo vision only or laser scanners. Reactive algorithms can include basic intelligence like finding the way back to the original path after an avoidance maneuver. The other way is to map obstacles and to plan a safe flight trajectory through these maps. Common approaches use e.g. force fields, vector field histograms or velocity obstacles. Contrary to reactive approaches, map-based avoidance algorithms can include advanced path planning. This can be realized by considering the flight envelope, using a-priori knowledge, learning information, finding a way through a labyrinth and flying an optimized trajectory to a desired target without collision. An autonomous vehicle must be able to operate in unknown environments, and an obstacle map has to be created with the help of sensors before a path planning algorithm can be applied. A procedure of creating obstacle maps from a UAV is the main focus of this paper. According to the literature grid maps are commonly used as they allow an easy fusion with data from different sensors including noise reduction and simultaneous pose estimation. But they require a lot of memory space. Further, they do not separate single objects. Since obstacle avoidance is often based on metric polygonal or topological world models, grid maps must be converted into other map types. The presented approach creates an occupancy grid at first and creates a polygonal map out of it. This polygonal representation is used to avoid obstacles later on. The mapping algorithm is tested on an unmanned helicopter.


    Access

    Access via TIB

    Check availability in my library


    Export, share and cite



    Title :

    A system for vision-based flights in unknown urban environments


    Contributors:


    Publication date :

    2008


    Size :

    7 Seiten, 18 Bilder, 20 Quellen


    Type of media :

    Conference paper


    Type of material :

    Storage medium


    Language :

    English




    Autonomous Tail-Sitter Flights in Unknown Environments

    Lu, Guozheng / Ren, Yunfan / Zhu, Fangcheng et al. | ArXiv | 2024

    Free access

    A vision and GPS-based real-time trajectory planning for MAV in unknown urban environments

    Flores, Gerardo / Zhou, Shuting / Lozano, Rogelio et al. | IEEE | 2013


    Vision-Based Flexible Fleet Maneuvering in Unknown Cluttered Environments

    Wang, Yuanzhe / Wang, Danwei | Springer Verlag | 2022


    Autonomous vision-based helicopter flights through obstacle gates

    Andert, Franz / Adolf, Florian-M. / Goormann, Lukas et al. | Tema Archive | 2010