Abstract: The present disclosure provides systems and methods that employ tolerance values defining a level of vehicle control precision for motion control of an autonomous vehicle. More particularly, a vehicle controller can obtain a trajectory that describes a proposed motion path for the autonomous vehicle. A constraint set of one or more tolerance values (e.g., a longitudinal tolerance value and/or lateral tolerance value) defining a level of vehicle control precision can be determined or otherwise obtained. Motion of the autonomous vehicle can be controlled to follow the trajectory within the one or more tolerance values (e.g., longitudinal tolerance value(s) and/or a lateral tolerance value(s)) identified by the constraint set. By creating a motion control framework for autonomous vehicles that includes an adjustable constraint set of tolerance values, autonomous vehicles can more effectively implement different precision requirements for different driving situations.

Abstract: Systems and methods for controlling an autonomous vehicle are provided. In one example embodiment, a computer-implemented method includes obtaining data representing a first trajectory including one or more states corresponding to a motion path. The method includes determining a second trajectory based at least in part on the first trajectory, the second trajectory including a first state corresponding to the motion path, and one or more secondary states corresponding to the motion path, the one or more secondary states indicating a state of the autonomous vehicle relative to the first state. The method includes determining one or more control signals based at least in part on the second trajectory. The method includes controlling a motion of the autonomous vehicle according to the motion path, based at least in part on the one or more control signals.

Abstract: A LIDAR calibration module can detect a set of return signals from a plurality of fiducial targets for a set of laser scanners of an autonomous vehicle's LIDAR module. The autonomous vehicle can rest on an inclined platform of a rotating turntable to increase range variation in the return signals. Based on the set of return signals, the LIDAR calibration system can generate a set of calibration transforms to adjust a set of intrinsic parameters of the LIDAR module.

Abstract: A LIDAR calibration system can detect a first set of return signals from a plurality of fiducial targets in a calibration facility for a lower set of laser scanners of the LIDAR module. The LIDAR calibration system can also detect a second set of return signals from one or more planar surfaces associated with a calibration trigger location on a road network for an upper set of laser scanners of the LIDAR module. Based on the first and second sets of return signals, the LIDAR calibration system can generate a set of calibration transforms to adjust a set of intrinsic parameters of the LIDAR module.

Abstract: Systems and methods are directed to providing speed limit context awareness during operation of an autonomous vehicle. In one example, a computer-implemented method for applying speed limit context awareness in autonomous vehicle operation includes obtaining, by a computing system comprising one or more computing devices, a plurality of features descriptive of a context and a state of an autonomous vehicle. The method further includes determining, by the computing system, a context response for the autonomous vehicle based at least in part on a machine-learned model and the plurality of features, wherein the context response includes a derived speed constraint for the autonomous vehicle. The method further includes providing, by the computing system, the context response to a motion planning application of the autonomous vehicle to determine a motion plan for the autonomous vehicle.

Abstract: The present disclosure provides systems and methods that enable an autonomous vehicle motion planning system to learn to generate motion plans that mimic human driving behavior. In particular, the present disclosure provides a framework that enables automatic tuning of cost function gains included in one or more cost functions employed by the autonomous vehicle motion planning system.

Abstract: An on-demand transportation management system can collect historical data of harmful events of human-driven vehicles (HDVs) operating throughout a given region. For each road segment of the given region, the system can determine a fractional risk value for the HDVs, and based on the fractional risk value for each road segment, the system can route drivers within the given region along lowest risk route options.

Abstract: An on-demand transportation management system can receive transport requests in connection with an on-demand transportation service, each transport request indicating a start location and a destination. The system can determine a set of candidate vehicles to service each transport request, and can further determine a non-trip risk value for servicing the transport request. The system may then select an optimal vehicle in the set of candidate vehicles based at least in part on the non-trip risk value.

Abstract: An on-trip monitoring system for an on-demand transportation service can monitor live log data from autonomous vehicles (AVs) operating throughout a given region.

Abstract: An autonomous vehicle software management system can distribute AV software versions to safety-driven autonomous vehicles (SDAVs) operating within a given region. The system can receive log data from the SDAVs indicating any trip anomalies of the SDAVs while executing the AV software version. When a predetermined safety standard has been met based on the log data, the system can verify the AV software version for execution on fully autonomous vehicles (FAVs) operating within the given region.

Abstract: A risk regression system can computationally determine aggregate risk values for each of a plurality of routes for each transport request of an on-demand transportation service. The aggregate risk values can be based on fractional risk quantities determined through, for example, autonomous vehicle logs form autonomous vehicles operating throughout a given region. Based on the aggregate risk values, the risk regression system can facilitate vehicle matching between autonomous vehicles and human-driven vehicles.

Abstract: An on-demand transportation management system can collect vehicle fleet utilization data corresponding to human-driven vehicles (HDVs) and autonomous vehicles (AVs) operating within a given region in connection with an on-demand transportation service. The on-demand transportation management system can then establish a set of selection priorities for respective areas of the given region based on the vehicle fleet utilization data, each selection priority indicating whether a respective area of the given region is to favor HDVs or AVs for servicing transport requests.

Abstract: A transportation management system can maintain a set of driver logs for drivers operating throughout a given region, where each driver log indicates driving characteristics of a respective driver. The system can determine a destination for the respective driver operating a vehicle from an initial location to the destination, and determine a set of routes between the initial location and the destination. Based at least in part on the driving characteristics of the respective driver, the system can determine an individualized risk value for the respective driver for each route of the set of routes, and select an optimal route from the set of routes based, at least in part, on the individualized risk value for each route.

Abstract: An on-demand transportation management service can perform a selection process between a set of safety-driven autonomous vehicles (SDAVs), fully autonomous vehicles (FAVs), and human-driven vehicles (HDVs) to service transport requests based on a variety of selection parameters.

Abstract: An autonomous vehicle (AV) software management system can collect historical data of harmful events of human-driven vehicles (HDVs) within an autonomy grid on which AVs operate. For each path segment of the autonomy grid, the system can determine a fractional risk value for HDVs. The system may also receive AV data from a fleet of AVs operating throughout the autonomy grid, and for each path segment of the autonomy grid, the system can evaluate AV performance against the fractional risk values for HDVs based on the received AV data.

Abstract: An on-demand transportation management system can receive transport requests from requesting users for an on-demand transportation service for a given region, each transport request indicating a pick-up location and a destination. The system can determine a candidate set of vehicles, within a proximity of the pick-up location, to service each transport request. The system may then determine an individual risk value for each vehicle in the candidate set of vehicles for servicing the transport request, based, at least in part, on the individual risk value for each vehicle of the candidate set of vehicles, the system can select a vehicle from the candidate set of vehicles to service the transport request.

Abstract: An on-trip monitoring system for an on-demand transportation service can monitor trips performed by AVs throughout a given region, each trip including a given destination. For each on-trip AV, the system can determine a set of post-trip options for the on-trip AV, and select a most optimal post-trip option from the set of post-trip options. The system may then transmit a post-trip command instructing the on-trip AV to execute the most optimal post-trip option after arriving at the given destination.

Abstract: An on-demand transportation management system can perform trip classification operations based at least on risk by establishing a set of risk thresholds for safety-driven autonomous vehicles, fully autonomous vehicles, and/or the software versions executable by such vehicles. Based on the trip classifications, the on-demand transportation management system can perform vehicle matching operation with requesting users.

Abstract: An on-trip monitoring system for an on-demand transportation service can receive log data from autonomous vehicles operating throughout a given region. Based on a set of triggering conditions, the on-trip monitoring system can transmit switching commands to cause the AVs to switch between software versions and/or operative modes.

Abstract: A computer system maintains, for a given geographic region, a data structure that identifies a traction value for each of a plurality of locations of a road network within a geographic region. At least of a start or destination location is determined for the trip. The computer system may plan the trip, including selecting at least one of a route or a vehicle to use for the trip, based on the traction values of one or more of the plurality of locations.