LiDAR Alignment Optimization

Automated calibration technique for LIDAR systems on mobile robots that leverages the robot's mobility to gather data for calibration. The technique involves capturing LIDAR measurements while the robot spins at different locations with varying distances to a calibration target. These measurements are processed to determine calibration data for aligning the LIDAR units. This allows accurate localization and avoids misalignment issues when using fixed-plane LIDARs.

Source

Automatic calibration of multiple lidar sensors on autonomous vehicles using lane lines. The method involves using lane lines detected by the lidars to calibrate the relative positions and orientations of the lidars. By finding common lane line features visible to multiple lidars, geometric constraints can be derived to determine the lidar-to-lidar and lidar-to-vehicle transformations. This allows flexible, scene-agnostic calibration without markers or prior models.

Source

Method for calibrating multiple lidar sensors using deep learning and point cloud recognition to quickly and accurately extract inter-sensor calibration matrices. The method involves finding common objects in the lidar point clouds using a neural network, extracting those object areas, and applying scan matching to precisely calculate the calibration matrices. It provides a faster, more precise, and automated way to calibrate lidar sensors compared to manual methods.

Source

Calibrating LiDAR sensors mounted on a vehicle that have non-overlapping fields of view (FOVs) to accurately represent the environment scanned by multiple sensors. The calibration involves collecting sensor data from the non-overlapping LiDARs in a calibration environment, aligning the data in a global coordinate system, and iteratively aligning the data from each sensor until the aligned sensor views match. This extrinsically calibrates the sensors without assuming overlap. Intrinsic calibration further refines the sensor parameters. Validation checks calibration accuracy over time.

Source

Calibrating a LiDAR sensor on a moving vehicle without requiring specific trajectories or high-precision inertial navigation systems. The method involves calculating the relative position and orientation of the LiDAR between point clouds as the vehicle travels. This provides the necessary orientation data to determine the calibration parameters from the LiDAR coordinate system to the vehicle forward coordinate system. The method improves convenience and accuracy of LiDAR calibration by leveraging the vehicle's normal motion instead of specialized maneuvers.

Source

Calibrating the relative position between point clouds collected from multiple 3D LiDAR sensors in an autonomous navigation system. The method involves extracting planes from the point clouds, finding corresponding planes between sensors using similarity, initializing extrinsic parameters, and then optimizing them to minimize measurement point variance. The plane-based approach allows calibration using environmental features without special objects or trajectory estimation.

Source

Calibrating the external parameters between a lidar sensor and an inertial measurement unit (IMU) in an autonomous navigation system. The calibration method involves optimally solving a factor graph-based model to estimate the rotation and translation relationship between the lidar and IMU coordinate systems. It separates the calibration from the odometer, allows simultaneous rotation and translation estimation, and avoids plane feature requirements.

Source

Calibrating multiple LiDAR sensors with non-overlapping fields of view on an autonomous vehicle. The calibration method involves using a vehicle turntable system to transition the sensors from an uncalibrated state to a calibrated state. For intrinsic calibration, the sensors collect data from a calibration environment and optimize a model to estimate the intrinsic parameters like offsets. For extrinsic calibration, the sensors align sweeps using inter-sweep alignment and then globally align using a pose graph. Periodic calibration validation is also performed.

Source

Validating the calibration of LiDAR sensors mounted on vehicles to ensure accurate perception and navigation. The method involves collecting sensor data from the LiDAR in a calibration environment, segmenting the point cloud into planes, and measuring normal distances between the sweeps and planes. A validation score is determined using these distances. If the score is below a threshold, the calibration is validated. If not, it indicates a calibration issue. This allows proactive calibration maintenance and remediation.

Source

Efficient lidar calibration method for autonomous vehicles that avoids the need for complex mechanical devices. The method uses point cloud data captured from the lidar at a target position to calculate the external parameters. The initial point cloud is transformed through matrices based on target parameters, centroid differences, and homogeneous transformations to generate the lidar's translational and rotational freedom relative to the vehicle's coordinate system.

Source

Automatic calibration of multi-laser lidars on autonomous vehicles like AGVs without the need for manual calibration. The calibration involves calculating the relative poses between the lidars using the difference in calculated poses when they are stationary. This allows the lidars to self-calibrate without needing accurate measurement of their installation poses. By using the same map for all lidars, the relative poses can be obtained through navigation calculations.

Source

Calibrating RGB-D cameras and lidar sensors together for mobile robot perception systems. The calibration involves extracting feature points on a calibration plate using both sensors, and then iteratively refining the external parameter matrix between the two coordinate systems using anisotropic error models for lidar and RGB-D cameras. This accounts for the inherently anisotropic error distributions of each sensor type. The calibration improves the accuracy of converting depth measurements from both sensors into a common coordinate system for better fusion and perception.

Source

Diagonally installed dual lidar calibration method for autonomous vehicles with diagonally mounted lidars. The calibration involves capturing data from the lidars while moving in a straight line. The lidar points and angles are calculated from the known vehicle position and motion. This data is used to determine the actual lidar offsets from the nominal installed positions. The offsets are then applied to correct the lidar data for accurate mapping and navigation. The method addresses the challenge of calibrating diagonally mounted lidars with minimal overlap by using the vehicle motion to triangulate the lidar positions.

Source

A method for calibrating the external parameters between lidar and inertial measurement units (IMU) in autonomous vehicles, to improve the accuracy of sensing data fusion. The calibration uses a feature-based approach that extracts distinctive points from the lidar point clouds and aligns them with the IMU data. This avoids the need for specialized equipment or indoor environments. The method involves: 1) Identifying feature points in the lidar point clouds using a feature extraction algorithm. 2) Extracting corresponding points from the IMU data using a projection method. 3) Matching the lidar and IMU points based on their feature descriptions. 4) Estimating the external parameters (e.g., position and orientation) between the lidar and IMU using the matched points. This enables calibration using lidar data alone, without additional equipment or environments.

Source

Calibration method for lidar and camera on autonomous vehicles to improve perception accuracy after assembly. The method involves aligning calibration points generated by scans from the lidar and images from the camera. The 3D points are converted to 2D camera coordinates. Coordinate errors are computed for the 2D points and features. Initial transformation matrix is optimized based on these errors to accurately represent the lidar-camera relative position/attitude. This joint calibration resolves coordinate discrepancies between sensors after vehicle assembly.

Source

Rapid, accurate joint calibration method for multiple lidars and cameras in unstructured scenes without using calibration plates or structured features. The method involves adjusting the relative positions and heights of the bracket and sensor mounts to align the sensor fields of view. This allows joint calibration of the sensors without relying on external reference points. The bracket design enables rapid, on-the-fly calibration of multiple lidars and cameras mounted on unmanned vehicles in real-world scenarios without requiring pre-calibration or structured environments.

Source

A method for joint calibration of external parameters of multiple sensors like lidar, depth camera and RGB camera used in autonomous driving. The calibration is done using a device with a calibration plate and markings. The sensors simultaneously collect point cloud, depth and RGB data of the calibration device. The calibration plate center is calculated in each sensor's coordinate system using the markings. This allows calculating the calibration between any two sensors without resetting the device or using different calibration methods.

Source

Method for accurately calibrating the external parameters between lidar and position/orientation sensors in autonomous vehicles. The calibration improves accuracy in environments with limited plane features and provides 6D accuracy compared to 5D for previous methods. The calibration involves extracting planar features from lidar point clouds, projecting them into world coordinates, finding corresponding points across frames, and optimizing parameters using voxel grids and ground reference points.

Source

Hybrid calibration method for lidar sensors in autonomous driving to improve accuracy and efficiency compared to traditional calibration methods. The method involves converting point cloud data from both lidars to a common coordinate system, calibrating each lidar separately, then adjusting their parameters using a hybrid algorithm. This leverages the strengths of both lidars and avoids manual intervention.

Source

A method and system for calibrating multiple lidars and integrated navigation systems that improves accuracy and versatility compared to existing methods. The calibration method uses a combined approach that does not require additional inputs or preparation. It estimates initial external parameters between each lidar and integrated navigation system, then adjusts them based on factors like clock differences and vehicle vibrations. This provides calibrated parameters between each lidar and integrated navigation system.

Source