研究目的
To develop a decentralized framework for collaborative localization of heterogeneous systems using a modified MSCKF approach with range-based collaboration and environment constraints.
研究成果
The proposed collaborative MSCKF framework effectively enables decentralized 3D localization for heterogeneous robot teams without external computation, maintaining the original MSCKF properties while incorporating range-based updates and environment constraints. Experimental results demonstrate its validity and potential for applications like collaborative mapping and surveillance, with suggestions for future work on scalability and optimization.
研究不足
The framework may have computational inefficiencies with small range measurement windows, and the PDF truncation method can be sensitive to the number of samples, potentially requiring importance sampling for robustness. It is tested in specific indoor and outdoor environments, and generalization to larger teams or different sensor setups is not fully explored.
1：Experimental Design and Method Selection:
The methodology involves designing a two-level collaborative MSCKF filter that integrates range measurements and environment constraints using a truncated unscented Kalman filter. It builds on the original MSCKF framework for visual-inertial odometry and extends it to multi-robot scenarios without external computation.
2：Sample Selection and Data Sources:
Experiments use a Pioneer mobile robot equipped with a Bluefox camera, Xsens IMU, and UWB node, and Bebop2 UAVs with UWB nodes. Data includes IMU readings, camera images, and range measurements.
3：List of Experimental Equipment and Materials:
Equipment includes Pioneer robot, Bluefox camera, Xsens IMU, UWB nodes, Bebop2 UAVs, and Marvelmind ultrasonic system for ground truth.
4：Experimental Procedures and Operational Workflow:
The filter propagates IMU data, updates with camera features and range measurements, and incorporates environment constraints via PDF truncation. Data is shared between robots using ROS.
5：Data Analysis Methods:
Performance is evaluated using RMSE for position and rotation errors, and convergence is analyzed based on range window size and constraint inclusion.
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
