Abstract: In one embodiment, sensor data are collected from one or more sensors mounted on an autonomous driving vehicle (ADV) while the ADV is moving within a region of interest (ROI) that includes a number of obstacles. The collected sensor data are operated on to obtain obstacle data associated with the obstacles, location data, and a number of timestamps that correspond to the obstacle data and the location data. For each of the timestamps, positions of the obstacles are mapped to some of the obstacle data that correspond to the timestamp based on the location data, thereby generating mapped information of the obstacles. The mapped information is automatically labelled to generate labelled data, where the labelled data is utilized to subsequently train a machine learning algorithm to recognize obstacles during autonomous driving of an ADV.

Abstract: A sensor aggregation framework for autonomous driving vehicles is disclosed. In one embodiment, sensor data is collected from one or more sensors mounted on an autonomous driving vehicle (ADV) while the ADV is moving within a region of interest (ROI) that includes a number of obstacles. The sensor data includes obstacle information of the obstacles and vehicle data of the ADV. Each of the vehicle data is timestamped with a current time at which the vehicle data is captured to generate a number of timestamps that correspond to the vehicle data. The obstacle information, the vehicle data, and the corresponding timestamps are aggregated into training data. The training data is used to train a set of parameters that is subsequently utilized to predict at least in part future obstacle behaviors and vehicle movement of the ADV.

Abstract: A system adjusts a vehicle speed prior to a lane change for an autonomous driving vehicle (ADV). The system: plans a hypothetical rough path without lane change; determines virtual non-blocking Station-Time (ST) boundaries of potentially interfering vehicles in a target lane, comprising shifting the hypothetical rough path to an ST coordinate system associated with the target lane; determines projected ST boundaries, comprising projecting the virtual non-blocking ST boundaries onto a real ST coordinate system associated with a current lane; determines residual ST boundaries, comprising keeping only part of the projected ST boundaries where t>t0, wherein t0 is a parameter; and performs speed optimization based on the residual ST boundaries.

Abstract: According to some embodiments, a system operates an ADV. In one embodiment, the system perceives a driving environment surrounding the ADV based on sensor data obtained from a plurality of sensors, including perceiving a moving obstacle that is moving relative to the ADV. The system projects the moving obstacle as a figure onto a station-time (ST) coordinate system, wherein the ST coordinate system indicates a distance between the figure and a reference point at different points in time. And the system, for each of a plurality of predetermined processing time intervals, determines two points of the figure in the ST coordinate system to represent a shape of the figure, wherein the shape of the figure is utilized to plan a trajectory to drive the ADV to avoid colliding with the moving obstacle.

Abstract: System and method for training a machine learning model are disclosed. In one embodiment, for each of the driving scenarios, responsive to sensor data from one or more sensors of a vehicle and the driving scenario, driving statistics and environment data of the vehicle are collected while the vehicle is driven by a human driver in accordance with the driving scenario. Upon completion of the driving scenario, the driver is requested to select a label for the completed driving scenario and the selected label is stored responsive to the driver selection. Features are extracted from the driving statistics and the environment data based on predetermined criteria. The extracted features include some of the driving statistics and some of the environment data collected at the different points in time during the driving scenario.

Abstract: A learning based speed planner for autonomous driving vehicles (ADV) is disclosed. An ADV is set into human-driving mode. Driving control elements are under control of a human driver, and other ADV logic is enabled. The ADV plans a route path on a segment of the route having an obstacle. ADV logic generates a station-time graph for the path of the segment, and a grid of cells to encompass the path and obstacle. A feature vector is generated from the grid. Human driving behavior is recorded as the ADV is navigated along the path. Recorded driving data for a large plurality of paths, obstacles and ADVs is transmitted to a server to generate a speed model. The speed model is downloaded to one or more ADVs for use in autonomous driving mode, to determine an initial speed to use in similar driving situations.

Abstract: A new cost design is disclosed for evaluating candidate path curves for navigating an autonomous driving vehicle (ADV) through a segment of a route which may include an obstacle. Each point on each candidate path curve has a plurality of attributes having logical values and an associated priority of evaluation, and at least one numeric attribute having an associated priority of evaluation. A cost for each path curve is determined using the attributes and priorities, and a least cost path curve is selected using the attributes and priorities. By comparing attribute values in accordance with priority, and utilizing logical values, the efficiency of determining path curve cost and selecting a least cost path curve is substantially improved.

Abstract: In one embodiment, a system receives vehicle information from one or more ADVs. The system determines a location and a heading of each ADV from the vehicle information of the ADV. For each of the ADVs, the system determines if the ADV is approaching a traffic light junction based on the location and the heading of the ADV. The system sends the vehicle information of the ADVs to a traffic light control system in response to determining the ADV is approaching the traffic light junction, where the vehicle information is used by the traffic light control system to direct a traffic flow at the traffic light junction, including adjusting a time duration of a light signal at one or more traffic lights disposed at the traffic light junction in advance of the ADVs arriving at the traffic light junction.

Abstract: A perception module is configured to perceive a driving environment surrounding an autonomous driving vehicle (ADV) based on sensor data, and to generate perception information using various perception models or methods. The perception information describes the perceived driving environment. Based on the perception information, a planning module is configured to plan a trajectory representing a route or a path for a current planning cycle. The ADV is then controlled and driven based on the trajectory. In addition, the planning module determines a critical region (also referred to as a critical area) surrounding the ADV based on the trajectory in view of a current location or position of the ADV. The metadata describing the critical region is transmitted to the perception module via an application programming interface (API) to allow the perception module to generate perception information for a next planning cycle in view of the critical region.

Abstract: In one embodiment, a system of an ADV perceives a driving environment surrounding the ADV using a plurality of sensors mounted on the ADV. The system identifies a blind spot based on the perceived driving environment surrounding the ADV. The system in response to identifying the blind spot, receives an image having the blind spot from an image capturing device disposed within a predetermined proximity of the blind spot. In some embodiments, the system receives the image having the blind spot from a remote server communicatively coupled to the image capturing device. The system identifies an obstacle of interest at the blind spot of the ADV based on the image. The system generates a trajectory based on the obstacle of interest at the blind spot to control the ADV to avoid the obstacle of interest.

Abstract: In one embodiment, a system sends current location information of the ADV to an alert service over a network, where the alert service is communicatively coupled to a number of ADVs. The system receives a broadcasted alert signal from the alert service, where the alert service has determined that the ADV is or will be located in an alert area, and the location of the alert area is determined based on a location of a dispatched vehicle having a higher priority of traveling. In response to receiving the broadcast alert signal, the system examines a current state and the current location of the ADV in view of the alert area to determine whether the ADV should overtake or yield the alert area based on a set of rules. The system generates a trajectory to control the ADV to navigate the alert area based on the examination.

Abstract: According to some embodiments, a system performs an emergency stop when a speed planning optimization fails to generate a speed curve. In one embodiment, in response to an emergency stop request, the system generates one or more path-time analytical curves, where each of the one or more path-time curves is represented by a polynomial function. The system selects one of the path-time analytical curves to determine whether the selected path-time analytical curve satisfies a set of evaluation criteria. If the selected path-time analytical curve does not satisfy the set of evaluation criteria, the system selects a next one of the path-time analytical curves for evaluation. If the selected path-time analytical curve satisfies the set of evaluation criteria, the system generates a trajectory based on the selected path-time analytical curve to control the ADV during an emergency stop.

Abstract: A real-time perception adjustment and correction and driving adaption method for autonomous driving is provided. The system will analyze the driving behaviors of surrounding vehicles surrounding an ADV, which is utilized to improve its original perception based on the analysis and to adapt to the updated driving environment. In one embodiment, in addition to the perception information provided by the sensors, the system analyzes the behaviors of the surrounding vehicles based on the perception information. Based on the behaviors of the surrounding vehicles, the system may detect there is may be an obstacle that has not been detected based on the perception information. Alternatively, the system may detect that an obstacle determined based on the perception information actually may not exist based on the behaviors of the surrounding vehicles. The paths created for the ADV may then adjusted accordingly to improve the autonomous driving of the ADV.

Abstract: In one embodiment, a system is designed to determine the requirement of a perception range for a particular type of vehicles and a particular planning and control technology. A shadow filter is used to connect a scenario based simulator and a PnC module, and tuning the parameters (e.g. decreasing the filter range, tuning the probability of obstacles to be observed among frames) of shadow filter to mimic the real world perceptions with a limited range and reliabilities. Based on the simulation results (e.g., a failure rate, smoothness, etc.), the system is able to determine the required perception distance for the current PnC module. A PnC module represents a particular autonomous driving planning and control technology for a particular type of autonomous driving vehicles. Notice that the PnC module is replaceable so that this method is suitable for different PnC algorithms representing different autonomous driving technologies.

Abstract: A station-time (S-T) graph may be obtained in response to a first reference line representing a path from a first location to a second location associated with an autonomous driving vehicle (ADV). One or more kernels may be applied to the S-T graph. Each of the one or more kernels may indicate a plurality of points on the S-T graph. One or more constraints may be applied to the S-T graph. Each of the one or more constraints may indicate a condition for points in the S-T graph. A set of speeds for portions of the path is determined based on the one or more kernels and the one or more constraints.

Abstract: A first reference line representing a route from a first location to a second location associated with an autonomous driving vehicle (ADV). In response to the first reference line multiple sets of sample points are generated based on the first reference line. Each set of sample points includes a first subset of sample points and a second subset of sample points that are offset from the first subset of sample points. A plurality of segments is generated between the multiple sets of sample points. A path for the ADV is generated based on the plurality of segments. The path includes one sample point from each of the multiple sets of sample points.

Abstract: A method efficiently determines whether a trajectory of an obstacle to an autonomous driving vehicle (ADV) will require navigation adjustment for the ADV. The trajectory of the obstacle is represented by an ordered plurality of points, {P0 . . . PM}, and a reference line of the ADV is represented by an ordered plurality of points, {R0 . . . RN}, N>M. If the obstacle and ADV are heading in the same direction, then, for a first point P0 on the obstacle trajectory, the ADV finds a point S0 in {R0 . . . RN} that is the least distance from P0 to the reference line. For each of the remaining obstacle points, Pi, the search for a least distance point, Si, is limited to the portion of the reference line Si?1 to RN. If the obstacle and ADV are heading in opposing directions, the search process can be performed from PM, toward P0, in a similar fashion.

Abstract: In one embodiment, when an ADV is driving on a trajectory generated based on a reference line, a separate processing thread is executed to precalculate a new reference line as a future reference line for a future planning cycle in parallel. The future reference line is being created while the ADV is moving along a trajectory generated based on the original reference line and before reaching a location corresponding to the starting point of the future reference line. The future reference line is overlapped with an end section of the original reference line, such that the future reference line can be connected to the end section of the original reference line. The future reference line serves an extension of the original reference line before the ADV reaches the end section of the original reference line.

Abstract: Via a first processing thread, an ADV is controlled according to a first trajectory that was generated based on a first reference line starting at a first location. Concurrently via a second processing thread, a second reference line is generated based on a second location of the first trajectory that the ADV will likely reach within a predetermined period of time in future. The predetermined period of time is greater than or equals to an amount of time to generate a reference line for the ADV. The second reference line is generated while the ADV is moving according to the first trajectory and before reaching the second location. Subsequently, in response to determining that the ADV is within a predetermined proximity of the second location, a second trajectory is generated based on the second reference line without having to calculate the second reference line at the second location.

Abstract: According to some embodiments, a system calculates a first trajectory based on a map and a route information. The system performs a path optimization based on the first trajectory, traffic rules, and an obstacle information describing obstacles perceived by the ADV. The path optimization is performed by performing a spline curve based path optimization on the first trajectory, determining whether a result of the spline curve based path optimization satisfies a first predetermined condition, performing a finite element based path optimization on the first trajectory in response to determining that the result of the spline curve based path optimization does not satisfy the first predetermined condition, performing a speed optimization based on a result of the path optimization, and generating a second trajectory based on the path optimization and the speed optimization to control the ADV.