| Summary: | 碩士 === 國立成功大學 === 電機工程學系 === 104 === Simultaneous localization and mapping is an important function for home service robots. Generally, if the environment map is known, then self-localization for a robot is easy to implement. Likewise, if the robot has precise location of its position, establishing the environment map would be facile. However, simultaneous localization and mapping (SLAM) is like the chicken and egg conundrum, the robot has to build map and localize its position at the same time. This thesis presents Piecewise Linear Feature based SLAM (PLFSLAM) method that includes Rao-Blackwellised particle filter (RBPF), piecewise linear feature extraction, and scan matching algorithm. In PLFSLAM, piecewise linear features are extracted as the map features, and each particle carries an individual map. Besides, it adds scan matching algorithm that can reduce the number of particles to achieve accurate estimation of the robot position and build compact map with lower memory consumption. There are many applications of SLAM, in this thesis, we utilize it to implement the accompanying walk with human being. Accompanying walk has become a popular application in human-robot interaction. As known the environment map, the robot can plan a walking path while maintain the relative position with human. A RGB-D sensor is combined with laser information to track the person, and the Taylor series expansion (TSE) velocity estimator is used to estimate the velocity and to predict the pose in next time step of the person. The experimental results demonstrate the accuracy of the proposed PLFSLAM and the effectiveness of accompanying walk with human being.
|
|---|
