Abstract: Relocating and/or re-sizing map elements using an updated pose graph without introducing abnormalities to the map data may comprise determining a transformation between a source node of a first pose graph and a target node of a second pose graph and determining a modification to a map element based at least in part on the transformation. The techniques may include determining a stress on the map based at least in part on one or more modifications to map elements and determining if the stress meets or exceeds a threshold. In instances where the stress meets or exceeds a threshold, a modification may be altered, reversed, and/or indicated in a notification transmitted to a user interface.

Abstract: Techniques for representing a scene or map based on statistical data of captured environmental data are discussed herein. In some cases, the data (such as covariance data, mean data, or the like) may be stored as a multi-resolution voxel space that includes a plurality of semantic layers. In some instances, individual semantic layers may include multiple voxel grids having differing resolutions. Multiple multi-resolution voxel spaces may be merged or aligned to generate combined scenes based on detected voxel covariances at one or more resolutions.

Abstract: Techniques for integrating sensor data into a scene or map based on statistical data of captured environmental data are discussed herein. The data may be stored as a multi-resolution voxel space and the techniques may comprise first applying a pre-alignment or localization technique prior to fully integrating the sensor data.

Abstract: Systems and methods for calibrating multiple inertial measurement units on a system include calibrating a first of the inertial measurement units relative to the system using a first calibration model, and calibrating the remaining inertial measurement unit(s) relative to the first inertial measurement unit using a second calibration model. The calibration of the remaining inertial measurement unit(s) to the first inertial measurement unit can be based on a rigid body model by aligning a rotational velocity of the first inertial measurement unit with a rotational velocity of the remaining inertial measurement unit(s).

Abstract: Techniques described are related to determining when a discrepancy between data of multiple sensors (e.g., IMUs) might be attributable to a sensor error, as opposed to operating conditions, such as sensor bias or noise. For example, the sensor data is passed through one or more filters (e.g., bandpass filter) that model the bias or noise, and the filtered data may then be compared for consistency. In some examples, consistency may be based on residuals or some other metric describing discrepancy among the filtered sensor data.

Abstract: A soft-constraint technique for refining an initial pose graph may eschew using a hard constraint that identifies different sensor data and/or poses as necessarily being associated with a same portion of an environment. Instead, the soft-constraint technique may employ a loss function with a convergence basin that may be defined based at least in part on an object classification that strongly penalizes candidate locations within a distance associated with the convergence basin. These candidate locations may be based at least in part on one or more object detections associated (1:1) with one or more poses of the initial pose graph. This may result in one or more candidate locations that do not merge with other candidate locations, giving the pose graph optimization the permissiveness or softness according to the techniques described herein.

Abstract: Techniques for using a set of variables to estimate a vehicle velocity of a vehicle are discussed herein. A system may determine an estimated velocity of the vehicle using a minimization based on an initial estimated velocity, steering angle data and wheel speed data. The system may then control an operation of the vehicle based at least in part on the estimated velocity.

Abstract: Techniques for utilizing microphone or audio data to detect and responding to low velocity impacts to a system such as an autonomous vehicle. In some cases, the system may be equipped with a plurality of microphones that may be used to detect impacts that fail to register on the data captured by the vehicle's inertial measurement units and may go undetected by the vehicle's perception system and sensors. In one specific example, the perception system of the autonomous vehicle may identify a period of time in which a potential low velocity impact may occur. The autonomous vehicle may then utilize the microphone or audio data associated with the period of time to determine if an impact occurred.

Abstract: Techniques are described for determining whether a point cloud registration (e.g., alignment) between two sets of data points is valid. Such techniques can include determining voxelized representations of the sets of data points, and comparing characteristics of spatially aligned voxels within the voxelized representations. Characteristics of voxels to be compared can include classification labels of data points associated with voxels, including whether or not voxels correspond to free space. Point cloud registrations determined to be invalid can be given a weighting to be used in a subsequent high definition (HD) map building process. Generated maps can then be deployed for use in autonomous vehicles.

Abstract: Techniques associated with generating and maintaining sparse geographic and map data. In some cases, the system may maintain a factor graph comprising a plurality of nodes. In some cases, the nodes may comprise pose data and sensor data associated with an autonomous vehicle at the geographic position represented by the node. The nodes may be linked based on shared trajectories and shared sensor data.

Abstract: Techniques are described for using variables associated with vehicle wheels (e.g., linear velocity at a wheel and orientation of the wheel) to estimate velocity of a vehicle during a turn maneuver. In examples of the disclosure, in association with one or more wheels, a wheel orientation during the maneuver and a linear speed during the maneuver may be determined, and well as a yaw rate (e.g., from an inertial measurement unit, gyroscope, etc.) of the vehicle. Examples of the present disclosure include, based on the variables associated with the wheel(s) and the yaw rate associated with the turn maneuver, estimating a vehicle velocity, which may be used by various downstream components, such as to determine or update a pose of a vehicle as part of a localization operation.

Abstract: The techniques discussed herein include modifying a Kalman filter to additionally include a loss component that dampens the effect measurements with large errors (or measurements indicating states that are rather different than the predicted state) have on the Kalman filter and, in particular, the updated uncertainty and/or updated prediction. In some examples, the techniques include scaling a Kalman gain based at least in part on a loss function that is based on the innovation determined by the Kalman filter. The techniques additionally or alternatively include a reformulation of a Kalman filter that ensures that the uncertainties determined by the Kalman filter remain symmetric and positive definite.

Abstract: A computer-implemented method. Includes obtaining pointwise data indicating, for a plurality of time steps, a pointwise measurement of a state of an object detected by an object detection system. Includes obtaining, from a runtime model, runtime data indicating, for the plurality of time steps, a runtime estimate of the state of the object. Includes processing, by a benchmark model, the pointwise data to determine, for the plurality of time steps, a benchmark estimate of the state of the object. Includes evaluating a metric measuring, for the plurality of time steps, a deviation between the runtime estimate and the benchmark estimate of the state of the object. Includes updating, based on the on the evaluation of the metric, the runtime model.

Abstract: Techniques for using a set of non-steering variables to estimate an angle of a wheel are described. For example, a yaw rate, a linear velocity of a wheel, and vehicle dimensions (e.g., offset between the wheel and a turn-center reference line), can be used to estimate the angle of the wheel. Among other things, estimating angles based on non-steering variables may provide redundancy (e.g., when determined in parallel with steering-based command angles or other commanded angles) and/or may be used to validate commanded angles based on steering components.

Abstract: A computer-implemented method. Includes obtaining pointwise data indicating, for a plurality of time steps, a pointwise measurement of a state of an object detected by an object detection system. Includes obtaining, from a runtime model, runtime data indicating, for the plurality of time steps, a runtime estimate of the state of the object. Includes processing, by a benchmark model, the pointwise data to determine, for the plurality of time steps, a benchmark estimate of the state of the object. Includes evaluating a metric measuring, for the plurality of time steps, a deviation between the runtime estimate and the benchmark estimate of the state of the object. Includes updating, based on the on the evaluation of the metric, the runtime model.

Abstract: Techniques for, among other things, detecting faults associated with inertial measurement units (IMUs) of a vehicle when multiple IMUs are coupled to the vehicle are described herein. The techniques may include receiving first data from a first IMU of the vehicle and receiving second data from a second IMU of the vehicle. Based at least in part on the first data and the second data, a rotation of the first IMU relative to the second IMU may be calculated. The calculated rotation between the first IMU and the second IMU may be indicative of a fault associated with the first IMU or the second IMU. In response to detecting the fault, an action may be performed with respect to the first IMU or the second IMU to correct for the fault.

Abstract: Techniques are described for determining whether a point cloud registration (e.g., alignment) between two sets of data points is valid. Such techniques can include determining voxelized representations of the sets of data points, and comparing characteristics of spatially aligned voxels within the voxelized representations. Characteristics of voxels to be compared can include classification labels of data points associated with voxels, including whether or not voxels correspond to free space. Point cloud registrations determined to be invalid can be given a weighting to be used in a subsequent high definition (HD) map building process. Generated maps can then be deployed for use in autonomous vehicles.

Abstract: Techniques associated with generating and maintaining sparse geographic and map data. In some cases, the system may maintain a factor graph comprising a plurality of nodes. In some cases, the nodes may comprise pose data and sensor data associated with an autonomous vehicle at the geographic position represented by the node. The nodes may be linked based on shared trajectories and shared sensor data.

Abstract: Techniques for using a set of variables to estimate a vehicle velocity of a vehicle are discussed herein. A system may determine an estimated velocity of the vehicle using a minimization based on an initial estimated velocity, steering angle data and wheel speed data. The system may then control an operation of the vehicle based at least in part on the estimated velocity.

Abstract: Motion can be induced at a vehicle, e.g., by actuating components of an active suspension system, and first sensor data and second sensor data representing an environment of the vehicle can be captured at a first position and a second position, respectively, resulting from the induced motion. A second sensor can determine motion information associated with the first position and the second position. Calibration information about the sensor, the first sensor data, and the motion information can be used to determine an expectation of sensor data at the second position. A calibration error can be the difference between the second sensor data and the expected sensor data.