Abstract: Sensor data obtained from vehicles driving through a particular environment (e.g., a particular city or region being mapped) is used to identify and label traffic control features. Location, speed, and/or acceleration of mapping vehicles can be used to identify intersections that may have traffic control features. Environmental data, such as image data captured by one or more cameras, and point cloud data collected by lidar and/or radar sensors, is used to automatically detect traffic control features at the intersections.

Abstract: Planning, or path finding is one of many tasks performed by an autonomous vehicle (AV). Planning involves searching for and updating an optimal plan from a current pose to an end pose while avoiding obstacles. Planning can be particularly challenging in driving scenarios involving complex driving maneuvers, such as parking, making a U-turn, making a K-turn, etc. Searching for an optimal plan may take more time, e.g., a few seconds, in certain driving scenarios. Consumers of the plan may have a low tolerance for delay in receiving the optimal plan. A relay can be implemented to publish the latest plan to the consumers at a frequency. The relay can check whether the delay violates a dynamic threshold and raise an error accordingly. The searching process can be improved to reduce the delay in finding an optimal plan.

Abstract: Disclosed are embodiments for facilitating a multi-inertial measurement unit (IMU) combination unit. In some aspects, an embodiment includes receiving, at a multi-IMU combination unit (MICU), sensor data from a plurality of inertial data sensors of a same sensor type; for each inertial data sensor, calibrating and transforming the respective sensor data using a calibration estimate for the inertial data sensor, where the calibration estimate is based on pre-integration methods that provide individual kinematic feedback that is compared to fused kinematic feedback from a main filter; combining the calibrated and transformed sensor data from the plurality of inertial data sensors into a fused output for the same sensor type; sampling the fused output to provide a single inertial data measurement for the plurality of inertial data sensors; and providing the fused kinematic feedback for the calibration estimate, the fused kinematic feedback generated from the sampling of the single fused output.

Abstract: Disclosed are embodiments for facilitating real-time autonomous vehicle (AV) fleet parking availability. In some aspects, an embodiment includes determining, based on a first set of perception sensor inputs of an AV, sections of roadway that are used for vehicle parking; determining, based on a second set of perception sensor inputs, whether the sections of roadway are allowable for parking and are available for parking; transmitting parking location data comprising identification of parking location objects corresponding to the sections of roadway and signals indicating that the parking location object is used for vehicle parking, indicating whether the parking location object is allowable for parking, and indicating whether the parking location object is available for parking, wherein the parking location data is aggregated with other parking location data from other AVs into aggregated parking location data; and identifying a location for parking of the AV based on the aggregated parking location data.

Abstract: Platforms to support and manage an autonomous vehicle (AV) fleet can be implemented on and supported by cluster infrastructure. Cluster infrastructure may include different clusters, such as a cluster for an AV fleet with safety drivers, and a cluster for an AV fleet without safety drivers. Updates to the cluster infrastructure can be made to multiple clusters at once. Such updates may cause an outage that impacts multiple clusters and create a massive vehicle retrieval event for all AV fleets. To mitigate the risk and allow for prioritization of clusters, updates can be rolled out in a staged and ordered manner according to order values associated with different clusters. After applying an update to a cluster at a stage of the roll out, the cluster may be evaluated to confirm the success of the update, before the roll out can be moved to the next stage.

Abstract: Systems, methods, and computer-readable media are provided for maintaining cleanliness of an autonomous vehicle. In some examples, an autonomous vehicle fleet management device may receive cleanliness data corresponding to at least one autonomous vehicle. In some aspects, the autonomous vehicle fleet management device may determine, based on the cleanliness data, one or more parameters corresponding to at least one cleanliness issue associated with the at least one autonomous vehicle. In some cases, a remediation plan for the at least one cleanliness issue may be determined based on the one or more parameters. In some instances, at least one driving instruction may be sent to the at least one autonomous vehicle that is based on the remediation plan.

Abstract: Systems and methods for failover handling at a vehicle in response to a degraded state of operation are provided. For example, a method includes receiving, from a fallback vehicle path planner of a vehicle, data indicative of a fallback path; storing, at a memory of the vehicle, the data indicative of the fallback path; receiving, from a primary vehicle path planner of the vehicle, data indicative of a primary path; receiving an event trigger indicative of a degraded operating condition associated with the primary vehicle path planner; selecting, in response to the event trigger, the fallback path over the primary path; and sending, to a control system of the vehicle in response to the selecting, the stored data indicative of the fallback path.

Abstract: Disclosed are embodiments for facilitating autonomous vehicle simulated mileage data collection-guided operational design domain testing and test generation. In some aspects, an embodiment includes receiving performance results corresponding to simulated mileage accumulation of simulated autonomous vehicles (AVs); determining, from the performance results, performance metrics and corresponding confidence intervals for the performance metrics; identifying at least one of the performance metrics that does not satisfy a performance metric threshold or at least one of the corresponding confidence intervals that does not satisfy a confidence interval threshold; determining an identified domain corresponding to the at least one of the performance metrics or the at least one of the corresponding confidence intervals; and causing at least one of on-road supervised driving of one or more AVs to be initiated or one or more synthetic tests to be generated in accordance with the identified domain.

Abstract: Examples of the present disclosure provide a computer-implemented system, comprising instructions for performing operations including: estimating locations in a region of a map where features associated with water, snow or ice are likely to be formed under certain adverse weather conditions; simulating one of the adverse weather conditions; generating the features associated with the simulated one of the adverse weather conditions at the locations; determining a response of a perception stack of an autonomous vehicle (AV) to the adverse weather conditions observed at the locations; determining a reaction of the AV to the features generated in the region, the reaction of the AV being a function of a configuration of the AV; in response to the reaction, updating the configuration; repeating the determining the reaction and the updating the configuration until a desired reaction is obtained; and exporting a final configuration corresponding to the desired reaction to a physical AV.

Abstract: Systems and techniques are described for improving the accuracy of autonomous vehicle simulations by dynamically assigning friction coefficients to road features based on sensor data. A method of the disclosed technology can include steps for associating a predetermined friction coefficient to road features; receiving sensor data from a sensor on an autonomous vehicle; identifying, from the sensor data, the presence of a road feature; assigning in a simulation, a friction coefficient corresponding to the road feature based on the predetermined friction coefficient associated with road feature; and generating, based on the assigned friction coefficient, a simulation of a trip by the autonomous vehicle as the autonomous vehicle traverses the road feature.

Abstract: Systems and techniques are provided for rare scenario handling in autonomous vehicles. In some aspects, an AV management system can determine a target likelihood value range for identifying relevant rare scenarios. In some cases, the AV management system may use the target likelihood value range to generate new relevant rare scenarios that have a likelihood value that falls within the target likelihood value range. In some examples, the AV management system may initiate simulations of the generated relevant rare scenarios and capture synthetic AV scene data, which can be used to train machine learning models used in AVs. As a result, the performance of the AVs may be improved when faced with rare scenarios.

Abstract: Multi-sensor radar microDoppler holography is described. An example of a method includes performing a 2D Fourier transform calculation for multiple sets of raw Radar sensor data to generate Doppler spatial frequency map data, the raw sensor data including at least a first set of raw sensor data associated with a first Radar sensor at a first location and a second set of raw sensor data associated with a second Radar sensor at a second, different location; performing a DWT calculation of the Doppler spatial frequency map data; applying phase compensation to generate phase compensated data transforms; adding the phase compensated data transforms to generate a sum; performing an inverse 2D Fourier transform calculation to generate a transform result; and extracting microDoppler information and Doppler information extraction from the transform result.

Abstract: The subject disclosure relates to estimating a grade of a road and calibrating sensors of an autonomous vehicle based on the grade. A process of the disclosed technology can include effectuating an autonomous vehicle to perform one or more maneuvers that cause the autonomous vehicle to face different directions along a road, obtaining a first set of sensor measurements of an autonomous vehicle in a first pose, wherein the first set of sensor measurements includes a first measured specific force, obtaining a second set of sensor measurements of the autonomous vehicle in a second pose, wherein the second set of sensor measurements includes a second measured specific force, determining a direction of gravity based on the first measured specific force and the second measured specific force, and calibrating at least one sensor of the autonomous vehicle based on the determined direction of gravity.

Abstract: Build, execution, and testing on software may be performed remotely on cluster infrastructure. Tests can be scheduled on workers in the cluster infrastructure, and results of the tests are reported to the developer. If a test fails, the developer may receive an exit code from the failed test. For tests that are flaky, an exit code provides little to no benefit to identify the cause of the flakiness. To make it easier for the developer to debug flaky tests, a test identified to be flaky can be run many times in parallel, one or more parallel executions that result in failure can be paused, and one or more remote debugging sessions can be created for the one or more parallel executions. A developer can access the remote debugging session to inspect the paused execution of the test equipped with debugging tools to determine the cause of the flakiness.

Abstract: A method is described and includes acquiring sensor data produced by sensors of a plurality of vehicles traversing an area including a location of a user, wherein the vehicles traversing the area comprise a subset of a fleet of vehicles for providing rideshare services; processing the acquired sensor data to determine a category of the user; selecting a vehicle from the fleet of vehicles based on the category of the user, wherein the selected vehicle comprises at least one accommodation corresponding to the category of the user; and dispatching the selected vehicle to a pick-up location designated by the user.

Abstract: Systems and methods for dynamically suppressing noise at electric vehicle charging sites. In particular, systems and methods are provided for measuring the ambient noise at a charging site and adjusting electric vehicle chargers to reduce noise pollution. In some examples, electric vehicle charger cooling fans potentially generate a high level of noise, and the power output of a charger can be decreased to decrease the heat generated by the chargers, thereby reducing the need for fans. In various examples, reduction of noise pollution can be especially important in specified geographies (e.g., residential neighborhoods) and/or during selected timeframes (e.g., overnight). In various examples, the system interfaces with a central computer service (e.g., a dispatch service) to intelligently route autonomous vehicles to charging sites based on noise levels at various available charging sites.

Abstract: Systems and techniques are provided for expanding a scope of coverage of test scenarios for training an autonomous vehicle (AV). An example method can include identifying a maneuver of an AV; receiving, from a test repository, a plurality of tests that includes the maneuver; identifying one or more segments on a map of an operational design domain (ODD) that include a driving environment for the maneuver; determining a similarity between a driving scene of each of the plurality of tests and the one or more segments on the map of the ODD; and determining a degree of test coverage for each of the one or more segments for the maneuver based on the determined similarity.

Abstract: Systems and techniques are provided for simulating movement of an articulated vehicle. An example method can include generating a simulation environment for an autonomous vehicle, wherein the simulation environment includes at least one articulated vehicle that includes a tractor coupled to a first trailer using a first pivot joint; determining a first velocity of the first pivot joint based on a first simulated movement associated with the tractor; and determining a second simulated movement associated with the first trailer based on the first velocity of the first pivot joint and a first distance between the first pivot joint and a first point on the first trailer.

Abstract: The subject technology is related to autonomous vehicles (AV) and, in particular, to calibration of an sensors of an AV on open roads after calibrating the sensors in an undefined training area. An example method includes calibrating a plurality of sensors of the AV based on performing a plurality of training steps of increasing complexity within the undefined environment, wherein the plurality of sensors of the AV are uncalibrated with respect to each other, determining that the AV satisfies an open road requirement to navigate an open road based on completing the plurality of training steps in the undefined environment, and operating the AV on open roads subject to at least one constraint. The method of claim 1, wherein the at least one constraint comprises a list of prohibited maneuvers that limit operation of the AV on open roads. The constraints are configured to dynamically change.

Abstract: Systems and methods for measuring autonomous vehicle (AV) capabilities in a simulation environment are provided. A method includes selecting, by a computer-implemented system, a first plurality of sampling points in an N-dimensional parameter space associated with a vehicle capability test, classifying, by the computer-implemented system, each sampling point of the first plurality of sampling points into a first class based on a vehicle passing a test scenario associated with the respective sampling point; or a second class based on the vehicle failing the test scenario associated with the respective sampling point; and determining, by the computer-implemented system based on the classifying, a capability of the vehicle. The classifying can include computing a first hyperplane in the N-dimensional parameter space to separate the first plurality of sampling points into the first class and the second class.