SEROW IMU , Encoder , F / T RGB-D Image KFusion Visual Odometry KFusion Mapping Dense Map VO Pose Dense RGB

Visual Simultaneous Localization and Mapping (visual SLAM) constitutes a challenging task when applied to humanoid robots. While walking, the robot’s feet strike the ground and generate sudden accelerations due to rapid and sequential contact switching. These in turn give rise to visual motion blurriness that greatly compromises the performance of the system. In this work we present a dense visual SLAM approach that integrates information from IMU, robot kinematics and contact measurements, to overcome these issues.