Star-world navigation functions for convergence to a time-varying destination manifold

Date
2017
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Robot navigation and motion planning are important tasks in practical applications. Navigation functions is an important tool for that purpose. They serve as a potential field methodology for exact robot motion planning and control problems. However, traditional navigation functions are built for static environments, and existing navigation functions with time-varying destinations only apply to sphere worlds. This dissertation proposes an extension to navigation function methodologies that applies to star worlds with time-varying destinations. Based on the existing literature of time-varying navigation functions in sphere worlds, time-varying diffeomorphic transformations are constructed, and a proof of the correctness of the overall construction that extends the time-varying navigation functions from sphere worlds to star worlds and star forests, is presented. A new environmental modeling method is also introduced, to reduce computational and analytical complexity of workspace modeling, without sacrificing too much in terms of representation generality. This environment modeling method provides unified expressions of cube-like obstacles in n-dimensional spaces. For the cube-like obstacles, analytical expressions of the implicit representations and the length of rays (segments from a "center" point in the interior to an arbitrary point on the boundary) are given. Thanks to the new environmental modeling method, real-time robot motion planning can be achieved with the proposed time-varying navigation functions, without elaborate workspace modeling and in a wide range of of complex enclosed environments. The proposed robot motion planning method is validated in both simulations and experiments. Calculation examples of time-varying navigation constructions are provided, and simulations with a ideal point robot without kinematics and dynamics constrains are conducted. Visualization of those time-varying potential fields with moving targets is also offered. Finally, experimental results with a real unicycle robot are demonstrated, achieving real-time control of robot target tracking tasks.
Description
Keywords
Applied sciences, Dynamic environment, Motion planning, Navigation function, Potential field, Robot navigation, Time-varying destination
Citation