Abstract: To operate an autonomous vehicle, a rail agent is detected in a vicinity of the autonomous vehicle using a detection system. One or more tracks are determined on which the detected rail agent is possibly traveling, and possible paths for the rail agent are predicted based on the determined one or more tracks. One or more motion paths are determined for one or more probable paths from the possible paths, and a likelihood for each of the one or more probable paths is determined based on each motion plan. A path for the autonomous vehicle is then determined based on a most probable path associated with a highest likelihood for the rail agent, and the autonomous vehicle is operated using the determined path.

Abstract: The technology relates to controlling a vehicle in an autonomous driving mode in accordance with behavior predictions for other road users in the vehicle's vicinity. In particular, the vehicle's onboard computing system may predict whether another road user will perform a “continuing” lane driving operation, such as going straight in a turn-only lane. Sensor data from detected/observed objects in the vehicle's nearby environment may be evaluated in view of one or more possible behaviors for different types of objects. In addition, roadway features, in particular whether lane segments are connected in a roadgraph, are also evaluated to determine probabilities of whether other road users may make an improper continuing lane driving operation. This is used to generate more accurate behavior predictions, which the vehicle can use to take alternative (e.g., corrective) driving actions.

Abstract: The technology relates to controlling a vehicle in an autonomous driving mode in accordance with behavior predictions for other road users in the vehicle's vicinity. In particular, the vehicle's onboard computing system may predict whether another road user will perform a “continuing” lane driving operation, such as going straight in a turn-only lane. Sensor data from detected/observed objects in the vehicle's nearby environment may be evaluated in view of one or more possible behaviors for different types of objects. In addition, roadway features, in particular whether lane segments are connected in a roadgraph, are also evaluated to determine probabilities of whether other road users may make an improper continuing lane driving operation. This is used to generate more accurate behavior predictions, which the vehicle can use to take alternative (e.g., corrective) driving actions.

Abstract: Jaywalking behaviors of vulnerable road users (VRUs) such as cyclists or pedestrians can be predicted. Location data is obtained that identifies a location of a VRU within a vicinity of a vehicle. Environmental data is obtained that describes an environment of the VRU, where the environmental data identifies a set of environmental features in the environment of the VRU. The system can determine a nominal heading of the VRU, and generate a set of predictive inputs that indicate, for each of at least a subset of the set of environmental features, a physical relationship between the VRU and the environmental feature. The physical relationship can be determined with respect to the nominal heading of the VRU and the location of the VRU. The set of predictive inputs can be processed with a heading estimation model to generate a predicted heading offset (e.g., a target heading offset) for the VRU.

Abstract: To operate an autonomous vehicle, a rail agent is detected in a vicinity of the autonomous vehicle using a detection system. One or more tracks are determined on which the detected rail agent is possibly traveling, and possible paths for the rail agent are predicted based on the determined one or more tracks. One or more motion paths are determined for one or more probable paths from the possible paths, and a likelihood for each of the one or more probable paths is determined based on each motion plan. A path for the autonomous vehicle is then determined based on a most probable path associated with a highest likelihood for the rail agent, and the autonomous vehicle is operated using the determined path.

Abstract: Aspects of the disclosure provide a method of identifying off-road entry lane waypoints. For instance, a polygon representative of a driveway or parking area may be identified from map information. A nearest lane may be identified based on the polygon. A plurality of lane waypoints may be identified. Each of the lane waypoints may correspond to a location within at least one lane. The polygon and the plurality of lane waypoints may be input into a model. A lane waypoint of the plurality of lane waypoints may be selected as an off-road entry lane waypoint. The off-road entry lane waypoint may be associated with the nearest lane. The association may be provided to an autonomous vehicle in order to allow the autonomous vehicle to use the association to control the autonomous vehicle in an autonomous driving mode.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium that generates path prediction data for agents in the vicinity of an autonomous vehicle using machine learning models. One method includes identifying an agent in a vicinity of an autonomous vehicle navigating an environment and determining that the agent is within a vicinity of a crossing zone, which can be marked or unmarked. In response to determining that the agent is within a vicinity of a crossing zone: (i) features of the agent and of the crossing zone can be obtained; (ii) a first input that includes the features can be processed using a first machine learning model that is configured to generate a first crossing prediction that characterizes future crossing behavior of the agent, and (iii) a predicted path for the agent for crossing the roadway can be determined from at least the first crossing prediction.

Abstract: Jaywalking behaviors of vulnerable road users (VRUs) such as cyclists or pedestrians can be predicted. Location data is obtained that identifies a location of a VRU within a vicinity of a vehicle. Environmental data is obtained that describes an environment of the VRU, where the environmental data identifies a set of environmental features in the environment of the VRU. The system can determine a nominal heading of the VRU, and generate a set of predictive inputs that indicate, for each of at least a subset of the set of environmental features, a physical relationship between the VRU and the environmental feature. The physical relationship can be determined with respect to the nominal heading of the VRU and the location of the VRU. The set of predictive inputs can be processed with a heading estimation model to generate a predicted heading offset (e.g., a target heading offset) for the VRU.

Abstract: Aspects of the disclosure provide a method of identifying off-road entry lane waypoints. For instance, a polygon representative of a driveway or parking area may be identified from map information. A nearest lane may be identified based on the polygon. A plurality of lane waypoints may be identified. Each of the lane waypoints may correspond to a location within at least one lane. The polygon and the plurality of lane waypoints may be input into a model. A lane waypoint of the plurality of lane waypoints may be selected as an off-road entry lane waypoint. The off-road entry lane waypoint may be associated with the nearest lane. The association may be provided to an autonomous vehicle in order to allow the autonomous vehicle to use the association to control the autonomous vehicle in an autonomous driving mode.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium that generates path prediction data for agents in the vicinity of an autonomous vehicle using one or more machine learning models. One of the methods includes identifying an agent in a vicinity of an autonomous vehicle navigating through an environment and determining that the agent is within a vicinity of a crossing zone across a roadway. The crossing zone can be a marked crossing zone or an unmarked crossing zone. For example, the crossing zone can be an unmarked crossing zone that has been identified based on previous observations of agents crossing the roadway.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for predicting near-curb driving behavior. One of the methods includes obtaining agent trajectory data for an agent in an environment, the agent trajectory data comprising a current location and current values for a predetermined set of motion parameters of the agent; processing a model input generated from the agent trajectory data using a trained machine learning model to generate a model output comprising a prediction of whether the agent will exhibit near-curb driving behavior within a predetermined timeframe, wherein an agent exhibits near-curb driving behavior when the agent operates within a particular distance of an edge of a road in the environment; and using the prediction to generate a planned path for a vehicle in the environment.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating unmapped U-turn predictions using a machine learning model. One of the methods includes obtaining features of an agent travelling on a roadway. One or more unmapped U-turn regions in a vicinity of the agent on the roadway are identified. For each of the unmapped U-turn regions and from at least the features of the agent, a respective likelihood score that represents a likelihood that the agent intends to make an unmapped U-turn at the unmapped U-turn region is generated. Based on the respective likelihood scores, one or more of the unmapped U-turn regions are selected. For each selected unmapped U-turn region, data specifying a candidate future trajectory in which the agent makes the unmapped U-turn at the selected unmapped U-turn region is provided as a possible future trajectory for the agent.

Abstract: The technology relates to controlling a vehicle in an autonomous driving mode in accordance with behavior predictions for other road users in the vehicle's vicinity. In particular, the vehicle's onboard computing system may predict whether another road user will perform a “continuing” lane driving operation, such as going straight in a turn-only lane. Sensor data from detected/observed objects in the vehicle's nearby environment may be evaluated in view of one or more possible behaviors for different types of objects. In addition, roadway features, in particular whether lane segments are connected in a roadgraph, are also evaluated to determine probabilities of whether other road users may make an improper continuing lane driving operation. This is used to generate more accurate behavior predictions, which the vehicle can use to take alternative (e.g., corrective) driving actions.

Abstract: To operate an autonomous vehicle, a rail agent is detected in a vicinity of the autonomous vehicle using a detection system. One or more tracks are determined on which the detected rail agent is possibly traveling, and possible paths for the rail agent are predicted based on the determined one or more tracks. One or more motion paths are determined for one or more probable paths from the possible paths, and a likelihood for each of the one or more probable paths is determined based on each motion plan. A path for the autonomous vehicle is then determined based on a most probable path associated with a highest likelihood for the rail agent, and the autonomous vehicle is operated using the determined path.

Abstract: Jaywalking behaviors of vulnerable road users (VRUs) such as cyclists or pedestrians can be predicted. Location data is obtained that identifies a location of a VRU within a vicinity of a vehicle. Environmental data is obtained that describes an environment of the VRU, where the environmental data identifies a set of environmental features in the environment of the VRU. The system can determine a nominal heading of the VRU, and generate a set of predictive inputs that indicate, for each of at least a subset of the set of environmental features, a physical relationship between the VRU and the environmental feature. The physical relationship can be determined with respect to the nominal heading of the VRU and the location of the VRU. The set of predictive inputs can be processed with a heading estimation model to generate a predicted heading offset (e.g., a target heading offset) for the VRU.

Abstract: Preventing fraud or misuse associated with payment instruments comprises a processor for training a machine-learning process based on historic data related to interactions of an instrument. The processor trains a machine-learning process based on historic data related to interactions of an instrument and instrument issuer with counter-parties and users. The processor receives a request to evaluate the instrument for a risk of fraud and enters the accessed data into the machine-learning process. The processor determines a first risk score based on the machine-learning process that is based on a likelihood that the instrument issuer will remit invoiced funds and a second risk score based on a likelihood that the instrument issuer will initiate chargebacks. The processor determines that a combination of the first and second risk score is higher than a configured threshold and instructs the requester not to interact with the instrument.

Abstract: Adding an instrument to a digital application comprises a processor for training a machine-learning process based on historic data related to interactions of an instrument with counter-parties and users. The processor receives a request to add the instrument to a digital application associated with a user and accesses data associated with the instrument, the user, and the digital application. The processor enters the accessed data into the machine-learning process and determines a risk score of adding the instrument to the digital application based on the machine-learning process. If the processor determines that the risk score is less than or equal to a configured threshold, the processor adds the instrument to the digital application. If the risk score is greater than the threshold, the processor rejects the addition of the instrument.

Abstract: Techniques for advertisement placement prioritization are described. In one or more implementations, a likely line of sight of a viewer of an event to a focus point of the event is determined. Then, a relative location of one or more display devices is identified in relation to the likely line of sight of the viewer. In addition, a priority level is assigned to respective display devices based on corresponding relative locations in relation to the likely line of sight. Further, various demographics and characteristics of audience members can be obtained to determine context-sensitive media content associated with the audience members to present via relatively higher priority display devices. Also, performances of event participants can be determined based on commentary analysis and used to determine context-sensitive media content associated with the event participants to present via the higher priority ad boards.

Abstract: Techniques for advertisement placement prioritization are described. In one or more implementations, a likely line of sight of a viewer of an event to a focus point of the event is determined. Then, a relative location of one or more display devices is identified in relation to the likely line of sight of the viewer. In addition, a priority level is assigned to respective display devices based on corresponding relative locations in relation to the likely line of sight. Further, various demographics and characteristics of audience members can be obtained to determine context-sensitive media content associated with the audience members to present via relatively higher priority display devices. Also, performances of event participants can be determined based on commentary analysis and used to determine context-sensitive media content associated with the event participants to present via the higher priority ad boards.