Keyframe-Based Tightly-Coupled SLAM with RGBD Camera and IMU

In this paper, we propose a novel inertial-assisted slam system intended for low-cost micro aerial vehicles (MAVs).The system sensor assembly consists of color camera, depth camera and a inertial measurement unit (IMU) with three-axis accelerometers/gyroscopes. We use IMU data to enable probability-based predetermined operations rather than hypothesis testing iterations to accelerate pose calculation and obtain its optimal estimation through nonlinear optimization.We illustrate the performance of our system by hovering a MAV in a GPS-denied environment which position accuracy can reach 3.9cm RMSE in a 189 second flight and its robustness is also illustrated in a complex indoor environment.