VILO SLAM: Tightly Coupled Binocular Vision-Inertia SLAM Combined with LiDAR.

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Existing visual-inertial SLAM algorithms struggle with accuracy and robustness in environments with sparse features or during constant velocity/pure rotation movements.
• Low-cost sensors often lead to performance degradation in challenging scenarios.

Purpose of the Study:

• To develop a tightly coupled Vision-IMU-2D Lidar Odometry (VILO) algorithm to address the limitations of current visual-inertial SLAM systems.
• To improve the accuracy and robustness of robot pose estimation, particularly in challenging environments.

Main Methods:

• A novel tightly coupled fusion of low-cost 2D lidar observations with visual-inertial data.
• Derivation of the Jacobian matrix for lidar residuals and construction of a vision-IMU-2D lidar residual constraint equation.
• Utilizing a nonlinear solution method for optimal robot pose estimation.

Main Results:

• The proposed VILO algorithm demonstrates reliable pose-estimation accuracy and robustness in various special environments.
• Significant reductions in position error and yaw angle error were observed compared to existing methods.
• The tightly coupled fusion approach effectively integrates 2D lidar data with visual-inertial information.

Conclusions:

• The VILO algorithm offers a significant improvement in accuracy and robustness for multi-sensor fusion SLAM.
• This approach enhances robot localization capabilities in visually challenging or dynamic scenarios.
• The tightly coupled fusion of 2D lidar, IMU, and vision data provides a robust solution for odometry.