研究目的
Enhancing the robustness of visual odometry (VO) in hostile environments by integrating MEMS-IMU to aid stereo-camera for robust pose estimation.
研究成果
The proposed fault-tolerant framework with stereo-camera and MEMS-IMU achieves robust and precise positioning information in hostile environments. It outperforms traditional VO and VIO methods in accuracy and robustness, with the ability to navigate properly even when stereo VIO system fails.
研究不足
The study focuses on hostile environments with sparse features, fast angular motions, or illumination changes. The performance may vary in other environments or with different sensor configurations.
1：Experimental Design and Method Selection:
The study employs a fault-tolerant adaptive extended kalman filter (FTAEKF) framework integrated with a stereo-camera and a MEMS-IMU for robust pose estimation in hostile environments. MEMS-IMU pre-integration is used to constrain feature points searching and matching, and to optimize the initial iterator pose.
2：Sample Selection and Data Sources:
Data is collected using a ZED stereo camera and a Xsens MTI-G-710 MEMS-IMU mounted on a tripod with three rollers, in hostile environments like a corridor and a tennis court.
3：List of Experimental Equipment and Materials:
ZED stereo camera (resolution: 1280 × 720, baseline: 12cm, frame rate: 15 HZ), Xsens MTI-G-710 MEMS-IMU (running at 200 HZ), NVIDIA Jetson TX2 processing platform, Novatel OEM6 GPS receiver.
4：Experimental Procedures and Operational Workflow:
The framework processes images and inertial measurements to estimate pose, using MEMS-IMU pre-integration to aid stereo VIO through constraining matching and predicting initial iteration pose. A fault-tolerant mechanism detects and isolates dramatic changes.
5：Data Analysis Methods:
The accuracy of the proposed algorithm is evaluated through RMSE with mark points and RTK references, comparing against ORB-SLAM2, MSF-EKF, and VINS-Mono.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
