Abstract: In one embodiment, a system generates a plurality of driving scenarios to train a reinforcement learning (RL) agent and replays each of the driving scenarios to train the RL agent by: applying a RL algorithm to an initial state of a driving scenario to determine a number of control actions from a number of discretized control/action options for the ADV to advance to a number of trajectory states which are based on a number of discretized trajectory state options, determining a reward prediction by the RL algorithm for each of the controls/actions, determining a judgment score for the trajectory states, and updating the RL agent based on the judgment score.

Abstract: In one embodiment, a method of training dynamic models for autonomous driving vehicles includes the operations of receiving a first set of training data from a training data source, the first set of training data representing driving statistics for a first set of features; training a dynamic model based on the first set of training data for the first set of features; determining a second set of features as a subset of the first set of features based on evaluating the dynamic model, each of the second set of features representing a feature whose performance score is below a predetermined threshold. The method further includes the following operations for each of the second set of features: retrieving a second set of training data associated with the corresponding feature of the second set of features, and retraining the dynamic model using the second set of training data.

Abstract: In one embodiment, a computer-implemented method of autonomously parking an autonomous driving vehicle, includes generating environment descriptor data describing a driving environment surrounding the autonomous driving vehicle (ADV), including identifying a parking space and one or more obstacles within a predetermined proximity of the ADV, generating a parking trajectory of the ADV based on the environment descriptor data to autonomously park the ADV into the parking space, including optimizing the parking trajectory in view of the one or more obstacles, segmenting the parking trajectory into one or more trajectory segments based on a vehicle state of the ADV, and controlling the ADV according to the one or more trajectory segments of the parking trajectory to autonomously park the ADV into the parking space without collision with the one or more obstacles.

Abstract: In one embodiment, a computer-implemented method of operating an autonomous driving vehicle (ADV) includes perceiving a driving environment surrounding the ADV based on sensor data obtained from one or more sensors mounted on the ADV, determining a driving scenario, in response to a driving decision based on the driving environment, applying a predetermined machine-learning model to data representing the driving environment and the driving scenario to generate a set of one or more driving parameters, and planning a trajectory to navigate the ADV using the set of the driving parameters according to the driving scenario through the driving environment.

Abstract: An autonomous driving vehicle (ADV) receives instructions for a human test driver to drive the ADV in manual mode and to collect a specified amount of driving data for one or more specified driving categories. As the user drivers the ADV in manual mode, driving data corresponding to the one or more driving categories is logged. A user interface of the ADV displays the one or more driving categories that the human driver is instructed collect data upon, and a progress indicator for each of these categories as the human driving progresses. The driving data is uploaded to a server for machine learning. If the server machine learning achieves a threshold grading amount of the uploaded data to variables of a dynamic self-driving model, then the server generates an ADV self-driving model, and distributes the model to one or more ADVs that are navigated in the self-driving mode.

Abstract: According to one embodiment, in response to a request to park an ADV into a parking lot, a remote server is accessed over a network (e.g., a VX2 link) to obtain a list of parking spaces that appear to be available in the parking lot. Based on the list of available parking spaces and the map associated with the parking lot, a route is generated to navigate through at least the available parking spaces. The ADV is driven according to the route to locate at least one of the available parking spaces and to park the ADV into the located available parking space. The centralized server is configured to periodically receive signals from a number of parking lots indicating which of the parking spaces of the parking lots are apparently available.

Abstract: In one embodiment, a data processing system for an autonomous driving vehicle (ADV) includes a processor, and a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations. The operations include generating a station-time (ST) graph based on perception data obtained from one or more sensors of the ADV, the ST graph including representing a location of an obstacle at different points in time, obtaining a tensor based on the ST graph, the tensor including a plurality of layers, the plurality of layers including a first layer having data representing one or more obstacles on a path in which the ADV is moving, applying a machine-learning model to the plurality of layers of the tensor to generate a plurality of numerical values, the plurality of numerical values defining a potential path trajectory of the ADV, and determining a path trajectory of the ADV based on the plurality of numerical values.

Abstract: In one embodiment, a method, apparatus, and system for trajectory planning for autonomous driving in an autonomous driving vehicle equipped with a low accuracy localization and perception module is disclosed.

Abstract: Methods and systems for operating an autonomous driving vehicle (ADV) are disclosed. A current state of an ADV is sampled at a first time to obtain a set of parameters. A cost, from a cost function that reflects desired control goals, is generated for a future time horizon based at least in part on the set of parameters. The cost is minimized with one or more constraints to obtain target control input values. For each of the target control input values, a lookup operation is performed using the control input value to locate a first mapping entry that approximately corresponds to the control input value. A first control command is derived from the first mapping entry. The ADV is controlled using the derived first control command.

Abstract: In one embodiment, a data processing system for an autonomous driving vehicle (ADV) includes a processor, and a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations. The operations include generating a station-time (ST) graph based on perception data obtained from one or more sensors of the ADV, the ST graph including representing a location of an obstacle at different points in time, obtaining a tensor based on the ST graph, the tensor including a plurality of layers, the plurality of layers including a first layer having data representing one or more obstacles on a path in which the ADV is moving, applying a machine-learning model to the plurality of layers of the tensor to generate a plurality of numerical values, the plurality of numerical values defining a potential path trajectory of the ADV, and determining a path trajectory of the ADV based on the plurality of numerical values.

Abstract: A sensor package, a sensor system, and a method for fabricating the sensor package are described that include a sensing chip having dispense chemistry disposed over an array of conductive elements. In an implementation, the sensor package may include a sensing chip that may include at least one conductive element, wherein the at least one conductive element may be part of an array of conductive elements defining a M by N matrix, where M is a number of rows of the at least one conductive element and N is a number of columns of the at least one conductive element. The sensing chip may further include dispense chemistry that may be disposed on the at least one conductive element and at least one contact pad. The sensor package may further include a microfluidic cap that may be positioned over at least a portion of the sensing chip, wherein the microfluidic cap and the sensing chip may define a cavity that may be configured to receive a fluid sample.

Abstract: In an embodiment, a learning-based dynamic modeling method is provided for use with an autonomous driving vehicle. A control module in the ADV can generate current states of the ADV and control commands for a first driving cycle, and send the current states and control commands to a dynamic model implemented using a trained neural network model. Based on the current states and the control commands, the dynamic model generates expected future states for a second driving cycle, during which the control module generates actual future states. The ADV compares the expected future states and the actual future states to generate a comparison result, for use in evaluating one or more of a decision module, a planning module and a control module in the ADV.

Abstract: An energy absorbing device for subway vehicle includes a movable anti-climber, a fixed anti-climber, an energy absorbing honeycomb, at least one collapse tube, two sliding-groove assemblies and two guide sliding rails. As the first stage energy absorbing unit of the total structure, the energy absorbing honeycomb directly bears the collision impact transferred from the movable anti-climber. Through the deformation of the energy absorbing honeycomb itself under pressure, the collision kinetic energy transfers into internal energy of deformation and heat, thus realizing the energy absorbing buffering. As the second stage energy absorbing buffering unit, the at least one collapse tube further absorbs the collision energy, thus further buffering and protecting the underframe of the vehicle body, as well as ensuring that the impact energy performs a multistage and serial operation according to a predetermined direction and sequence, thereby ensuring the reliability of operation of the energy absorbing device.

Abstract: Driving parameters (e.g., speed, heading direction) that an autonomous driving vehicle (ADV) likely utilize as target driving parameters are grouped into a number of ranges and one of the driving parameters in each range is selected as a driving parameter representative or a target driving parameter representing the respective range or segment. For each of the target driving parameters representing the ranges, a particle swarm optimization method is utilized to derive a set of most optimized coefficients for a controller (e.g., speed controller, steering controller) for controlling an ADV. A driving parameter to coefficient (parameter/coefficient) mapping table is generated to map a particular driving parameter representing a range of driving parameter to a set of one or more coefficients of a particular controller. The parameter/coefficient mapping table is utilized at real-time to configure a controller in response to a particular target driving parameter using the corresponding coefficients.

Abstract: According to some embodiments, a system selects a number of polynomials representing a number of time segments of a time duration to complete the path trajectory. The system selects an objective function based on a number of cost functions to smooth speeds between the time segments. The system defines a set of constraints to the polynomials to at least ensure the polynomials are smoothly joined together. The system performs a quadratic programming (QP) optimization on the objective function in view of the set of constraints, such that a cost associated with the objective function reaches a minimum while the set of constraints are satisfied. The system generates a smooth speed for the time duration based on the optimized objective function to control the ADV autonomously.

Abstract: According to some embodiments, a system determines a number of boundary areas having predetermined dimensions centered around each of a number of control points of a first reference line. The system selects a number of two-dimensional polynomials each representing a segment of an optimal reference line between adjacent control points. The system defines a set of constraints to the two-dimensional polynomials to at least ensure the two-dimensional polynomials passes through each of the boundary areas. The system performs a quadratic programming (QP) optimization on a target function such that a total cost of the target function reaches minimum while the set of constraints are satisfied. The system generates a second reference line representing the optimal reference line based on the QP optimization to control the ADV autonomously according to the second reference line.

Abstract: In one embodiment, a computer system generates a first vehicular path based on map information. The system collects data points representing geographical coordinates of vehicles that drove on a vehicular path lane at different points in time. The system segments the first vehicular path into path segments based on the collected data points. For each of the path segments, the system applies a smoothing function to select a subset of the data points that are within a predetermined proximity of the corresponding path segment and calculates a segment reference point to represent the path segment by combining the selected data points. The segment reference points of the path segments of the first vehicular path are interpolated to generate a second vehicular path such that the second vehicular path is used as a reference line to control ADVs driving on the first vehicular path.

Abstract: According to some embodiments, a system segments a first path trajectory selected from an initial location of the ADV into a number of path segments, where each path segment is represented by a polynomial function. The system selects an objective function in view of the polynomial functions of the path segments for smoothing connections between the path segments. The system defines a set of constraints to the polynomial functions based on adjacent path segments in view of at least a road boundary and an obstacle perceived by the ADV. The system performs a quadratic programming (QP) optimization on the objective function in view of the added constraints, such that an output of the objective function reaches a minimum. The system generates a second path trajectory representing a path trajectory with an optimized objective function based on the QP optimization to control the ADV autonomously.

Abstract: When generating a control command of an autonomous driving vehicle (ADV), a pitch status and/or a roll status of the road is determined. The control command is adjusted based on the pitch status and the roll status. For example, when an ADV is driving on an uphill or downhill road, a pitch status of the road is determined and a speed control command will be generated based on the pitch status of the road, such that the ADV have a similar acceleration rate as of driving on a flat road. Similarly, when the ADV is driving on a road that is tilted or rolled left or right, a roll status of the road is determined and a steering control command will be generated in view of the roll status of the road, such that the ADV have a similar heading direction as of driving on a flat road.

Abstract: In one embodiment, one or more first data items associated with a planned trip of a user riding in an autonomous driving vehicle (ADV) are displayed on a display device within the ADV. Each of the first data items is associated with a user selectable option to indicate whether the user wishes or allows the ADV to store each of the first data items in a persistent storage device. User inputs are received via a user interface such as touch screen of the display device, including a first selection indicating that the user wishes to store a first subset of the first data items. In response to the first selection, the first subset of the data items is stored in the persistent storage device of the ADV.