LIO-SAM is a tightly-coupled lidar-inertial odometry framework that achieves accurate, real-time mobile robot trajectory estimation and mapping. It uses a factor graph to incorporate multiple relative and absolute measurements, including loop closures, for state estimation. The framework estimates sensor motion using raw IMU data to de-skew point clouds and provides an initial guess for lidar odometry optimization. The estimated lidar odometry solution is used to estimate IMU bias. To ensure real-time performance, old lidar scans are marginalized for pose optimization, and scan-matching is performed at a local scale rather than a global scale. Keyframes are selectively added to the factor graph, and new keyframes are registered to a fixed-size set of prior sub-keyframes. The framework is evaluated on datasets from three platforms across various environments and shows improved accuracy and performance compared to existing methods like LOAM and LIOM. LIO-SAM is suitable for multi-sensor fusion and global optimization, and it incorporates GPS, compass, and altimeter data to eliminate drift. The system is efficient and can be deployed on low-power embedded systems. The results show that LIO-SAM achieves similar or better accuracy than LOAM and LIOM, with improved real-time performance and robustness in challenging environments.LIO-SAM is a tightly-coupled lidar-inertial odometry framework that achieves accurate, real-time mobile robot trajectory estimation and mapping. It uses a factor graph to incorporate multiple relative and absolute measurements, including loop closures, for state estimation. The framework estimates sensor motion using raw IMU data to de-skew point clouds and provides an initial guess for lidar odometry optimization. The estimated lidar odometry solution is used to estimate IMU bias. To ensure real-time performance, old lidar scans are marginalized for pose optimization, and scan-matching is performed at a local scale rather than a global scale. Keyframes are selectively added to the factor graph, and new keyframes are registered to a fixed-size set of prior sub-keyframes. The framework is evaluated on datasets from three platforms across various environments and shows improved accuracy and performance compared to existing methods like LOAM and LIOM. LIO-SAM is suitable for multi-sensor fusion and global optimization, and it incorporates GPS, compass, and altimeter data to eliminate drift. The system is efficient and can be deployed on low-power embedded systems. The results show that LIO-SAM achieves similar or better accuracy than LOAM and LIOM, with improved real-time performance and robustness in challenging environments.