Abstract: Techniques for determining whether to yield a vehicle to an oncoming object based on a stationary object are discussed herein. A vehicle can determine a drivable area in an environment through which the vehicle is travelling. A trajectory associated with the vehicle can be determined. A probability that the oncoming object will intersect a drive path associated with a vehicle trajectory can be determined. In some case, the probability can be based on various distance metric(s) (e.g., dynamic distance metrics) of the environment. If the oncoming object is determined to intersect the drive path, a stopping location for the vehicle can be determined to allow the oncoming object to traverse around the stationary object. The vehicle can be controlled based on sensor data captured by the vehicle and the probability.

Abstract: Controlling motion of an autonomous vehicle may comprise determining a state space representation of the environment associated with the autonomous vehicle based at least in part on sensor data. The autonomous vehicle may parameterize the state space according to arc length and lateral distance from a route reference. The autonomous vehicle may determine a state in the state space upon which to base trajectory generation (e.g., via a tree search, via a state space sampling technique based on cost) and may determine a set of control instructions (i.e., a trajectory) that would bring the autonomous vehicle to the arc length and lateral distance specified by the state. Determining the state for generating the trajectory may be based on a heuristic for determining approximately where the trajectory would bring the vehicle, since the arc length parameterized state space doesn't include an indication of location.

Abstract: Techniques for determining workstation devices and resource parameters for software tests are discussed herein. The workstation devices may include graphic processing units (GPUs) on which portions of the software tests are performed. The workstation devices and GPUs in the workstation devices may be allocated and designated for building and testing software programs. A task to be performed on a GPU cluster of a plurality of GPU clusters may be determined. The task may be added to a scheduling queue associated with the plurality of GPU clusters. The task may be assigned to an individual GPU, based at least in part on a memory constraint associated with the task. The task may be performed with the individual GPU of the GPU cluster to determine a result. The result may be returned to a computing device.

Abstract: Techniques for detecting and tracking objects in an environment are discussed herein. For example, techniques can include detecting a center point of a block of pixels associated with an object. Unimodal (e.g., Gaussian) confidence values may be determined for a group of pixels associated with an object. Proposed detection box center points may be determined based on the Gaussian confidence values of the pixels and an output detection box may be determined using filtering and/or suppression techniques. Further, a machine-learned model can be trained by determining parameters of a center pixel of the detection box and a focal loss based on the unimodal confidence value which can then be backpropagated to the other pixels of the detection.

Abstract: Techniques for determining environmental data of an environment associated with a vehicle are discussed herein. The vehicle can use the data to determine a vehicle configuration data, a trajectory for the vehicle to traverse, a route to a destination, or orientation for ingress/egress. The vehicle can determine control data based on the vehicle configuration data, the trajectory, or the route. The vehicle configuration data may be associated with a setting of a suspension component, a speed, or other component. The trajectory may be biased based on a source of the wind. The route may be selected from a first candidate route and a second candidate route using a first wind score associated with the first candidate route and a second score associated with the second candidate route. The vehicle can use a sensor, such as a wind sensor, to capture or otherwise determine the wind data.

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 determining unified futures of objects in an environment are discussed herein. Techniques may include determining a first feature associated with an object in an environment and a second feature associated with the environment and based on a position of the object in the environment, updating a graph neural network (GNN) to encode the first feature and second feature into a graph node representing the object and encode relative positions of additional objects in the environment into one or more edges attached to the node. The GNN may be decoded to determine a distribution of predicted positions for the object in the future. A predicted position of the object at a subsequent timestep may be determined by sampling from the distribution of predicted positions according to various sampling strategies. Alternatively, the predicted position of the object may be overwritten using a candidate position of the object.

Abstract: Techniques for correlating perception system errors with unwanted vehicle behavior. The techniques may include receiving log data associated with a vehicle traversing an environment. Based at least in part on the log data, an error associated with an output received from a perception component of the vehicle may be identified. The output of the perception component may be associated with a detection of an object in the environment. The techniques may also include determining that the error contributed to an unwanted behavior of the vehicle. In some examples, a simulation of a planner component of the vehicle may be ran using ground truth data to determine that the error contributed to the unwanted behavior. Based at least in part on the error contributing to the unwanted behavior, a parameter of the perception component may be updated to minimize the error and the unwanted behavior.

Abstract: An impact is detected with a vehicle having a body comprising parts which are deformable in an impact, by a secondary impact detection system configured to detect an impact that may not trigger a primary impact detection system. The secondary impact detection system comprises a plurality of deformation sensors attached to the deformable parts and configured to detect bending of the deformable parts. One or more processors are configured to receive a plurality of respective outputs from the plurality of deformation sensors, process the respective outputs to increase sensitivity to short-term changes relative to long-term changes and generate an indication of an impact dependent on detecting a processed output meeting a predetermined criterion. Outputs associated with a group of neighboring flex sensors may be used to localize the impact, characterize the source of impact, and/or reduce a likelihood of a false positive associated with the detection.

Abstract: Systems and techniques for detecting feature(s) and object(s) partially or completely obscured by stray light effects in the visible light spectrum are discussed herein. Non-visible light can be emitted from emitters. Light from an environment including visible and non-visible light can be captured at a sensing system via a lens system. The sensing system can filter the captured light using a dual band-pass filter to direct certain non-visible and visible light spectra to a beam splitter that then directs the non-visible light to a non-visible light sensor and the visible light to a visible light sensor. The sensor data generated by the non-visible light sensor can then be used to identify and classify feature(s) and object(s) that may be obscured by stray light effects in the data generated by the visible light sensor.

Abstract: Techniques for identifying and triaging issues associated with a vehicle in order to resolve the issues are described. The techniques may include receiving log data associated with a vehicle and generating an event timeline based at least in part on the log data. The event timeline may comprise a time-ordered record of events associated with operations of the vehicle. The techniques may also include identifying, based at least in part on an event included in the event timeline, a first occurrence of an issue associated with the vehicle. In some instances, a correlation may be determined between the first occurrence of the issue and a second occurrence of the issue, and a cause of the issue may also be determined based at least in part on the correlation. In some examples, the second occurrence of the issue may appear in the event timeline or another event timeline.

Abstract: A bidirectional vehicle may be capable of traveling in either of two directions and may change its direction of travel at any point for various reasons. In response to a change in the direction of travel, systems of a bidirectional vehicle may adjust one or more suspension components using various techniques to configure the vehicle with a higher ride frequency at the trailing axle than at the leading axle to enhance the ride quality and handling capabilities of the vehicle.

Abstract: Techniques for providing remote assistance to a vehicle are discussed. The techniques include receiving, from a vehicle, an indication of an event and displaying, on a display and to a user, a portion of an environment including the vehicle. The techniques further determine a valid region in the portion of the environment associated with a location at which the vehicle is capable of navigating. The techniques also display, on the display a footprint of the vehicle, where the footprint is associated with a position and orientation. The techniques further include transmitting the position and orientation of the footprint to the vehicle, which causes the vehicle to traverse in the environment to the position and orientation.

Abstract: Techniques predicting object trajectories and metadata are described herein. For example, the techniques may include inputting data representing a top-down view of an environment into a machine learned model. In some examples, the top-down view can include rasterized data or vectorized data. The machine learned model may output trajectories (e.g., position data, velocity data, acceleration data, etc.) and metadata (e.g., yaw, object state data, etc.) of one or more objects in the environment indicating future positions of the object(s). The machine learned model may rank the output trajectories and send information about the ranked output trajectories to a vehicle computing device for consideration during vehicle planning including simulation.

Abstract: Techniques for determining unified futures of objects in an environment are discussed herein. Techniques may include determining a first feature associated with an object in an environment and a second feature associated with the environment and based on a position of the object in the environment, updating a graph neural network (GNN) to encode the first feature and second feature into a graph node representing the object and encode relative positions of additional objects in the environment into one or more edges attached to the node. The GNN may be decoded to determine a predicted position of the object at a subsequent timestep. Further, a predicted trajectory of the object may be determined using predicted positions of the object at various timesteps.

Abstract: Techniques are discussed for interaction probabilities associated with regions of an environment around a vehicle. An interaction probability of a region may indicate a likelihood an object positioned at the region will interact with the vehicle. A top-down multi-channel image representing a top-down view of the environment and objects therein may be generated and input to a machine learned (ML) model. The ML model may output a probability map, a portion of the probability map comprising a region and an interaction probability associated with the region that indicates a likelihood objects positioned at the region will interact with the vehicle. A priority for resource assignment or analysis may be determined based on the interaction probability for an object positioned in the region. Control of the vehicle may be performed based at least in part on the priority for resource assignment or analysis.

Abstract: Techniques are discussed herein for generating, evaluating, and determining trajectories for autonomous vehicles traversing environments. A state transition model may be generated and used to determine a trajectory from multiple possible trajectories generated by one or more vehicle systems. In some examples, a state transition model may determine a trajectory based on the validation results of the possible trajectories, along with vehicle status data from one or more vehicle components. Various techniques described herein may improve vehicle safety and driving efficiency, by ensuring that the vehicle determines a safe and valid trajectory consistently while navigating the environment, while also being responsive to requests and status updates from various vehicle components.

Abstract: Techniques for clustering sensor data are discussed herein. Sensors of a vehicle may detect data points in an environment. Clustering techniques can be used in a vehicle safety system to determine connection information between the data points. The connection information can be used by a vehicle computing device that employs clustering and/or segmenting techniques to detect objects in an environment and/or to control operation of a vehicle.

Abstract: Techniques are discussed herein for executing log-based driving simulations to evaluate the performance and functionalities of vehicle control systems. A simulation system may execute a log-based driving simulation including playback agents whose behavior is based on the log data captured by a vehicle operating in an environment. The simulation system may determine interactions associated with the playback agents, and may convert the playback agents to smart agents during the driving simulation. During a driving simulation, playback agents that have been converted to smart agents may interact with additional playback agents, causing a cascading effect of additional conversions. Converting playback agents to smart agents may include initiating a planning component to control the smart agent, which may be based on determinations of a destination and/or driving attributes based on the playback agent.

Abstract: The techniques discussed herein may comprise an autonomous vehicle guidance system that generates a group of candidate trajectories based at least in part on sensor data. This group of candidate trajectories is clustered into two or more clusters and one or more representative trajectories may be determined for each cluster. The representative trajectory may be one of the trajectories in the cluster or a separately determined trajectory that represents the trajectories in a cluster, such as a mean or median trajectory. The representative trajectory may be used to predict how a dynamic object would react to the representative trajectory. This prediction may be used to determine potentially different costs associated with the different trajectories of the cluster with which the representative trajectory is associated. These costs may include costs to induce following behavior by the autonomous vehicle and/or cost(s) associated with conditionally available roadway portions.