CN109641589B

CN109641589B - Route planning for autonomous vehicles

Info

Publication number: CN109641589B
Application number: CN201780049995.2A
Authority: CN
Inventors: K·伊阿格内玛
Original assignee: Motional AD LLC
Current assignee: Motional AD LLC
Priority date: 2016-06-14
Filing date: 2017-06-13
Publication date: 2023-10-13
Anticipated expiration: 2037-06-13
Also published as: CN109641589A; WO2017218563A1; EP3468850A4; EP3468850A1

Abstract

Among other things, a determination is made as to: by time or time range, the autonomous vehicle is safely or robustly traveling the capabilities of the road feature or road segment or route being considered for the autonomous vehicle. The route conforms to the attributes of the stored road network information. If the computer has determined that the road feature or road segment or route cannot be safely or robustly traveled by the autonomous vehicle, the road feature or road segment or route is excluded from consideration. The determination by the computer is based on the attributes of the autonomous vehicle.

Description

Route planning for autonomous vehicles

Cross Reference to Related Applications

The application claims the benefits of the following U.S. applications: U.S. application Ser. No. 15/182,281 filed on day 2016, U.S. application Ser. No. 15/182,313 filed on day 2016, 6 and 14, U.S. application Ser. No. 15/182,360 filed on day 2016, 6 and 14, and U.S. application Ser. No. 15/182,400 filed on day 2016, each of which is incorporated herein by reference in its entirety.

Technical Field

The application relates to route planning for autonomous vehicles.

Background

The present description relates to route planning for autonomous vehicles.

The autonomous vehicle may be safely driven without human intervention during part of the journey or during the whole journey.

An autonomous vehicle includes sensors, actuators, computers, and communication devices to enable automatic generation and tracking of routes through an environment. Some autonomous vehicles have wireless two-way communication capability for: communication with a remotely located command center that can be manipulated by a human monitor, accessing data and information stored in a cloud service, and communication with an emergency service.

As shown in fig. 1, in typical use of an autonomous vehicle 10, a desired target location 12 (e.g., a destination address or street intersection) may be identified in various ways. The target location may be specified by an occupant (e.g., may be the owner of the vehicle or a passenger in a mobile-as-a-service (mobility-as-a-service) application). The target location may be provided by an algorithm (e.g., the algorithm may run on a central server in the cloud and be commissioned to optimize the task of location of the autonomous vehicle fleet in order to minimize the waiting time of the occupant when the occupant calls out the intelligent taxi). In some cases, the target location may be provided by a procedure (e.g., an emergency procedure that identifies the nearest hospital as the target location due to a medical emergency detected on the vehicle).

Given a desired target location, the routing algorithm 20 determines a route 14 through the environment from the current location 16 of the vehicle to the target location 12. We sometimes call this process "route planning". In some implementations, a route is a series of connected segments (which we sometimes refer to as road segments or segments for short) of roads, streets, and highways.

Routing algorithms typically operate by analyzing road network information. Road network information is typically a digital representation of the structure, type, connectivity, and other relevant information about the road network. A road network is typically represented as a series of connected road segments. The road network information may contain, in addition to identifying connectivity between road segments, additional information about the physical and conceptual attributes of each road segment including, but not limited to, geographic location, road name or number, road length and width, speed limit, direction of travel, lane edge boundary type, and any special information about the road segments, such as whether it is a bus lane, whether it is a right-turn lane or a left-turn lane only, whether it is part of a highway, a small road or a road, whether the road segment is allowed to stop or stand, and other attributes.

The routing algorithm typically identifies one or more candidate routes 22 from the current location to the target location. Identifying the best or optimal route 14 from among the candidate routes is typically accomplished by employing algorithms (such as the a, D, dijkstra algorithms, etc.) that identify routes that minimize the specified cost. The cost is typically based on one or more criteria, which typically include distance traveled along the candidate route, expected time to travel along the candidate route while taking into account speed constraints, traffic conditions, and other factors. The routing algorithm may identify one or more good routes to be presented to the occupant (or other person, e.g., an operator at a remote location) for selection or approval. In some cases, an optimal route may simply be provided to the vehicle track planning and control module 28, and the vehicle track planning and control module 28 has the function of directing the vehicle along the optimal route toward a target (which we sometimes refer to as a target location or simply a target).

As shown in fig. 2, road network information is typically stored in a database 30, the database 30 is maintained on a centrally accessible server 32, and the database 30 may be updated at a high frequency (e.g., 1Hz or higher). Network information may be accessed on demand (e.g., requested by the vehicle 34) or pushed to the vehicle by a server.

The road network information may have time information associated with the road network information for enabling a description of traffic regulations, parking regulations or other time-related effects (e.g., road segments where parking is not allowed during standard business hours or during weekends, for example), or for including information about expected travel times along the road segments at specific times of the day (e.g., during peak hours).

Disclosure of Invention

Generally, in an aspect, the feasibility of a route is determined. The route includes a sequence of connected road segments to be traveled by the autonomous vehicle from a starting location to a target location. The route conforms to the stored road network information.

Implementations may include one or a combination of two or more of the following features. The feasibility of the route includes the ability of the autonomous vehicle to safely travel the route. Feasibility includes that an autonomous vehicle cannot safely travel the route. The feasibility state of the route includes the ability of the autonomous vehicle to robustly travel the route. The feasibility state includes that the autonomous vehicle cannot robustly travel the route. The computer determines the feasibility of each route in the route set. The computer determines the feasibility up to a given time. The route is one of two or more candidate routes determined by the route planning process. The route is excluded from consideration as a candidate route. The feasibility state depends on the characteristics of the sensors on the vehicle. The characteristics include an actual performance level or an estimated performance level based on the current condition or a predicted future condition.

The feasibility state depends on the nature of the software process. The software process includes processing data from sensors on the vehicle. The software process includes motion planning. The software process includes decision making. The software process includes vehicle motion control. The characteristics include an actual performance level or an estimated performance level based on the current condition or a predicted future condition.

The feasibility state depends on the characteristics of the road feature. The characteristics of the road feature include spatial characteristics of an intersection, a roundabout (roundabout), or a highway entrance. Characteristics of road features include connectivity characteristics of intersections, roundabout, or highway exits. The characteristics of the road feature include spatial orientation. Characteristics of road features include road construction or traffic accidents. The characteristics of the road feature include roughness. Characteristics of road features include poor visibility due to curvature and slope. The characteristics of the road feature include poor previous drivability of the autonomous vehicle. The characteristics of the road feature include poor prior simulated performance of the model autonomous vehicle. Characteristics of road features include physical navigation challenges in severe weather.

Generally, in one aspect, a feasibility of one or more road segments belonging to a road network information body for travel of an autonomous vehicle is determined. The feasibility state of the road segment is made available for connection with road network information.

Implementations may include one or a combination of two or more of the following features. The feasibility of a road segment includes the ability of an autonomous vehicle to safely travel the road segment. Feasibility includes that an autonomous vehicle cannot safely travel the road segment. The feasibility of the route includes the ability of the autonomous vehicle to robustly travel the road segment. Feasibility includes that an autonomous vehicle cannot robustly travel the road segment. The computer determines the feasibility up to a given time. Feasibility depends on the characteristics of the sensors on the vehicle. The characteristics include an actual performance level or an estimated performance level based on the current condition or a predicted future condition.

Feasibility depends on the nature of the software process. The software process includes processing data from sensors on the vehicle. The software process includes motion planning. The software process includes decision making. The software process includes vehicle motion control. The characteristics include an actual performance level or an estimated performance level based on the current condition or a predicted future condition.

The feasibility state depends on the characteristics of the road segment. Characteristics of road segments include spatial characteristics of intersections, roundabout, or highway exits. Characteristics of road segments include connectivity characteristics of intersections, roundabout, or highway exits. Characteristics of the road segments include spatial orientation. Characteristics of road segments include road construction or traffic accidents. Characteristics of the road segment include roughness. Characteristics of road segments include poor visibility due to curvature and slope. The characteristics of the road segment include poor previous drivability of the autonomous vehicle. The characteristics of the road segments include poor prior simulated performance of the model autonomous vehicle. Characteristics of road segments include physical navigation challenges in inclement weather. The feasibility state of the route is updated as the autonomous vehicle is traveling the route.

Generally, in one aspect, a determination is made regarding: by time or time range, the autonomous vehicle is safely or robustly traveling the capabilities of the road feature or road segment or route being considered for the autonomous vehicle. The route conforms to the attributes of the stored road network information. If the computer has determined that the road feature or road segment or route cannot be safely or robustly traveled by the autonomous vehicle, the road feature or road segment or route is excluded from consideration. The determination by the computer is based on the attributes of the autonomous vehicle.

Implementations may include one or a combination of two or more of the following features. The attribute includes a characteristic of a sensor used by the autonomous vehicle during autonomous travel. Determination of the ability of an autonomous vehicle to safely or robustly travel is based on a software process that processes data representing the attributes of the autonomous vehicle. The attributes include performance levels of the sensors, including actual or estimated performance levels based on current or predicted future conditions. The attributes of the autonomous vehicle include the ability of one or more sensors to generate a data product of interest at a particular performance level. The attribute of the autonomous vehicle includes a failure mode of one or more sensors. The attributes of the autonomous vehicle include characteristics of the software process used by the autonomous vehicle. Characteristics of a software process include the ability of the software process to generate a data product of interest at a particular performance level. Characteristics of the software process include the ability of the data fusion network to generate data products of interest at a particular performance level. The software process includes a motion planning process. The software process includes a decision making process. The software process includes a motion control process.

Generally, in one aspect, a determination is made regarding: by time or time range, the autonomous vehicle is safely or robustly traveling the capabilities of the road feature or road segment or route being considered for the autonomous vehicle. The route conforms to the attributes of the stored road network information. If the computer has determined that the road feature or road segment or route cannot be safely or robustly traveled by the autonomous vehicle, the road feature or road segment or route is excluded from consideration. The determination by the computer is based on attributes of the environment in which the autonomous vehicle is traveling.

Implementations may include one or a combination of two or more of the following features. The environment includes road features. The attribute of the environment includes navigability of the autonomous vehicle. The attributes of the environment include spatial characteristics of the road features. The attributes of the environment include connectivity characteristics of the road features. Attributes of the environment include spatial orientation of road features. The attributes of the environment include the location of road construction or traffic accident. The attributes of the environment include road surface roughness of the road feature. The properties of the environment include a curved slope that affects visibility. The attributes of the environment include the nature of the marking of the road feature. The attributes of the environment include physical navigation challenges of road features associated with bad weather. The computer determines the ability of the autonomous vehicle to safely or robustly travel each road feature in the set of road features or each road segment in the set of road segments or each route in the set of routes.

The ability of an autonomous vehicle to safely or robustly travel a road feature or road segment or route depends on the characteristics of sensors on the vehicle. The characteristics include an actual performance level or an estimated performance level based on the current condition or a predicted future condition. The computer determines the ability of the autonomous vehicle to a given time. The route is one of two or more candidate routes determined by the route planning process. The ability of an autonomous vehicle to safely or robustly travel a road feature or road segment or route depends on the characteristics of the software process. The software process includes processing data from sensors on the vehicle. The software process includes motion planning. The software process includes decision making. The software process includes vehicle motion control. The characteristics include an actual performance level or an estimated performance level based on the current condition or a predicted future condition.

Generally, in one aspect, a determination is made regarding: by time or time range, the autonomous vehicle is safely or robustly traveling the capabilities of the road feature or road segment or route being considered for the autonomous vehicle. The route conforms to the attributes of the stored road network information. If the computer has determined that the road feature or road segment or route cannot be safely or robustly traveled by the autonomous vehicle, the road feature or road segment or route is excluded from consideration. The determination is based on an analysis of the performance of the autonomous vehicle.

Implementations may include one or a combination of two or more of the following features. The analysis of the performance of the autonomous vehicle includes previous driving performance associated with the road feature. The analysis of the performance of the autonomous vehicle includes previous simulated performance associated with road characteristics.

Generally, in an aspect, a route traveled by an autonomous vehicle up to a time or time range is selected from a group of two or more candidate routes, all candidate routes in the group having a feasibility state exceeding a feasibility threshold.

Implementations may include one or a combination of two or more of the following features. The feasibility state includes an indication of: the candidate route may be safely or robustly traveled by the autonomous vehicle or safely and robustly traveled by the autonomous vehicle. Information about a candidate route is received from a source as a feed of data products or data. The received information about the candidate route is part of road network information. The candidate route includes a route that includes at least one road segment that has not yet been validated.

These and other aspects, features, embodiments, and advantages, and combinations thereof, may be expressed as methods, systems, components, apparatus, program products, methods of doing business, apparatus or steps for performing functions, and in other ways.

Other aspects, features, embodiments, and advantages will become apparent from the following description and claims.

Drawings

FIG. 1 illustrates identification of a desired target location in typical use of an autonomous vehicle.

Fig. 2 shows a centrally accessible server for maintaining a database storing road network information.

Fig. 3 shows an example of the physical locations of the sensors and software processes.

Fig. 4 to 9 are schematic diagrams of road scenes.

FIG. 10 is a schematic diagram of a vehicle and a remotely located database.

Detailed Description

For route planning involving a human driven vehicle, it is generally assumed that the route from the current location to the target location identified by the routing algorithm is a route that can be safely driven by the driver, wherein the route is made up of connected road segments. However, for various reasons, this assumption may not be valid for the route identified by the routing algorithm for the autonomous vehicle. Due to the particular nature of road features and the ability of the vehicle to navigate certain road segments, intersections, or other geographic areas (which we will refer to broadly as road features) may not be able to safely navigate. Moreover, autonomous vehicles may not be able to safely navigate certain road features during certain periods of the day, during certain periods of the year, or under certain weather conditions.

Examples of physical locations of sensors in the vehicle and software processes at the cloud-based server and database are shown in fig. 3 and 10.

Sensor and software process

In many cases, such inability to safely navigate road features involves characteristics of sensors and software processes that the autonomous vehicle uses to sense the environment, process data from the sensors, learn about the conditions currently presented by the sensed environment and possibly the conditions presented by the sensed environment in the future, perform motion planning, perform motion control, and make decisions based on these senses and learnings. The following capabilities of the sensors and processes may be degraded or lost or unacceptably altered under certain conditions and at certain times, among others: sensing the environment, knowing the condition, performing motion planning and motion control, and making decisions.

Examples of such degradation or unacceptable changes in sensor and software process outputs are as follows:

sensor for sensing the environment of a vehicle

As shown on fig. 3, the following types of sensors 40 are generally available for vehicles (e.g., autonomous vehicles) having driver assistance capability or highly autonomous driving capability: sensors capable of measuring properties of the environment of the vehicle include, but are not limited to, for example, laser RADAR (LIDAR), RADAR (RADAR), monocular or stereoscopic video cameras in the visible, infrared or thermal spectrum, ultrasonic sensors, time of flight (TOF) depth sensors, and temperature and rain sensors, and combinations thereof. The data 42 from these sensors may be processed 44 to produce "data products" 46, for example, information regarding other vehicles, pedestrians, cyclists, scooters, cars, carts, animals, and other types, locations, speeds, and estimated future movements of the moving object. The data product also includes related objects and features such as static barriers (e.g., poles, signs, curbs, traffic marking cones and barrels, traffic signals, traffic signs, road dividers, and trees), road signs, and the location, type, and content of the road signs

The ability of software process 44 to use such sensor data to calculate such data products at a particular performance level depends on the properties of the sensor, such as detection range, resolution, noise characteristics, temperature dependence, and other factors. The ability to calculate these data products at a particular performance level may also depend on environmental conditions, such as the nature of the ambient lighting (e.g., whether direct sunlight, diffuse sunlight, sunrise or sunset conditions, dusk or darkness) exists, the presence of fog, mist, smoke or air pollution, whether it is raining or snowing or whether it has recently been raining or snowing, and other factors.

In general, it is possible to characterize the ability of a particular sensor (and associated processing software) to produce a data product of interest at a particular level of performance (e.g., a particular level of detection accuracy, detection range, ratio of true or false positives, or other metric) that is based on a measurable metric related to environmental conditions. For example, it is generally possible to characterize the range in which a particular monocular camera sensor can detect a moving vehicle at a particular performance level, which is a function of the ambient lighting levels associated with day and night conditions.

Furthermore, it is often possible to identify a specific failure mode of the sensor, i.e. a condition or environment in which the sensor will exactly deteriorate or cannot generate a data product of interest, and to identify a data product in which the sensor is not designed to be able to generate.

Fig. 9 shows an example of an autonomous vehicle sensor configuration.

Software for processing data from sensors

As described above, data from the sensors may be used by the software process 44 to generate various data products of interest. The ability of each of the software processes to generate data products that meet a specified performance level depends on the properties of the sensor software process (e.g., algorithm), which may limit their performance in scenes with certain properties, such as very high or very low densities of data features related to sensing tasks on hand.

For example, an algorithm for pedestrian detection that relies on data from a monocular vision sensor (we sometimes use the terms software process and algorithm interchangeably) may degrade or fail in its ability to detect more than a certain number of pedestrians at a specified performance level (e.g., a specified processing rate), and may therefore degrade or fail in scenes with a large number of pedestrians (in the sense that all pedestrians are not detected in scenes at the specified performance level). Furthermore, in a scenario with little geometric topography such as a flat parking lot, an algorithm for determining the position of a self (ego) vehicle (referred to as "locating") based on a comparison of lidar data collected from vehicle-mounted sensors with data stored in a map database may not be able to determine the current position of the vehicle at its ability to a specified level of performance (e.g., at a specified level of positional accuracy).

In general, the following capabilities are possible to characterize a particular sensor software process: the data product of interest is generated at a specific performance level according to the measurable scene attribute.

Typically, the data provided by the more than one sensor is combined in a data fusion framework implemented by one or more software processes in order to improve the overall performance of computing one or more data products. For example, data from a video camera may be combined with data from a lidar sensor to enable detection of pedestrians at performance levels designed to exceed those achievable by using the video camera or lidar sensor alone. In a data fusion scenario such as this, the above is still the case: it is generally possible to characterize the ability of a particular data fusion framework to produce a data product of interest at a particular level of performance.

Software process for motion planning

Vehicles capable of highly autonomous driving (e.g., autonomous vehicles) rely on a motion planning process, i.e., an algorithmic process that automatically generates and executes trajectories through the environment toward a specified short-term target. We use the term trajectory broadly to include paths from one place to another, for example. In order to distinguish the trajectory produced by the motion planning process from the route produced by the route planning process, we note that the trajectory is a path through the immediate surroundings of the vehicle (e.g. a distance scale is typically on the order of a few meters to hundreds of meters), which paths are specifically designed not to collide with the obstacle and generally have ideal characteristics related to: path length, ride quality, desired travel time, no violation of road rules, adherence to driving habits, or other factors.

Some motion planning processes employed on autonomous vehicles exhibit known limitations. For example, given that the vehicle is only moving in a forward direction but not in a backward direction, a certain motion planning procedure may be able to calculate a path for the vehicle from its current position to the target. Alternatively, a certain movement planning procedure may be able to calculate a path for the vehicle only when the vehicle is traveling at a speed less than a specified speed limit.

It is often possible to identify these and similar performance characteristics (e.g., limitations) on a motion planning process based on knowledge of the algorithm design of the process or its performance observed in simulation or experimental testing. Depending on the constraints of a particular motion planning process, it may be difficult or impossible to safely navigate in a particular area, such as a highway that needs to be driven at high speeds, or a multi-level parking structure that requires complex multi-point turns involving both forward and reverse maneuvers.

Software process for decision making

Vehicles capable of highly automated driving rely on a decision-making process, i.e., an algorithmic process, for automatically deciding on the proper short-term course of action for the vehicle at a given time, e.g., whether a stopped vehicle is exceeded or waiting behind it; whether to travel through a four-way stop intersection or to avoid (yield) vehicles that have previously reached the intersection.

Some decision-making processes employed on autonomous vehicles exhibit known limitations. For example, a certain decision making process may not be able to determine an appropriate course of action for a vehicle in certain scenarios of high complexity (e.g., in a roundabout including traffic lights) or in a multi-layer parking structure.

As in the case of motion planning processes, it is often possible to identify these and similar performance characteristics (e.g., limitations) on the decision-making process based on knowledge of the algorithm design of the process or its performance observed in simulation or experimental testing. Depending on the constraints on the particular decision making process, it may be difficult or impossible to safely navigate in a particular area.

Software process for vehicle motion control

Autonomous vehicles are generally intended to follow the trajectories provided by a motion planning process with a high degree of accuracy by employing a motion control process. The motion control process calculates a set of control inputs (i.e., steering, braking, and throttle inputs) based on analysis of the current deviation from the desired trajectory and predicted deviations, among other factors.

These motion control processes exhibit known limitations. For example, a certain motion control procedure may allow stable operation only in the forward direction, but not in the backward direction. Alternatively, a certain motion control process may have the ability to track a desired trajectory (to a specified accuracy) only when the vehicle is traveling at a speed less than a specified speed limit. Alternatively, a motion control process may have the following capabilities: steering or braking inputs requiring a certain level of lateral or longitudinal acceleration are only performed when the road surface friction coefficient exceeds a certain level.

As in the case of motion planning and decision making processes, it is often possible to identify these and similar limitations on the motion control process based on knowledge of the algorithmic design of the process or its performance observed in simulation or experimental testing. Depending on the constraints on the particular motion control process, it may be difficult or impossible to safely navigate in a particular area.

Safe or robust operation of the autonomous vehicle may be determined based on specific performance levels of sensors and software processes according to current and future conditions.

Characteristics of road features

The route planning process aims to exclude candidate routes as follows: including road features that may be determined as an autonomous vehicle unable to safely navigate. For this purpose, the route planning process may usefully consider a source of information particularly relevant to the autonomous vehicle, including information about characteristics of the road characteristics, such as spatial characteristics, orientation, surface characteristics, etc. Typically, this information will be used to avoid selecting the following routes for the autonomous vehicle: through areas of the road network where it is difficult or impossible for the vehicle to navigate at the required level of performance or safety. Examples of sources of information are described herein, as well as their interpretation of the effects on autonomous vehicle performance or safety.

Intersection, roundabout,Expressway entranceSpatial characteristics of other road features

As shown in the example shown in fig. 5, the road network information may contain or allow calculation of information about spatial characteristics of road intersections, roundabout, highway exits, or other road features including multi-lane surface roads and highways through a separate process. Such information may include, for example, the width of the road, the distance across the intersection (i.e., the distance from a point on the driving lane at the edge of the intersection to a point on the opposite lane at the opposite edge of the intersection), and the distance across the roundabout (i.e., the diameter of the roundabout)

In view of knowledge of the detected properties of the sensor system of the autonomous vehicle, analysis of such spatial characteristics may allow for determining that certain road segments are not navigable by the autonomous vehicle at a specified level of safety or robustness, regardless of or in view of a certain time or time of day or time range (e.g., after sunset and before sunrise). This may allow the autonomous vehicle to avoid, for example, certain intersections that are, for example, "too large to see both ends after sunset" in view of the actual limitations on the sensing capabilities of the autonomous vehicle and the allowable travel speed of the road. These limitations may make it impossible for autonomous vehicle sensors to provide data products to the motion planning process with sufficient time to react to impending traffic.

Intersection, roundabout,Expressway entranceConnectivity characteristics of other road features

As shown in the example shown in fig. 4, the road network information may contain or allow for calculation of information about connectivity characteristics of a particular road segment or individual road segment lanes or other road features by a separate process. For example, such information may include the orientation of intersecting road segments relative to each other. It may also include designations of specific travel lanes, such as right-turn only and left-turn only lane designations, or identification of highway entrance and exit ramps.

In view of the knowledge of the detection attributes of the autonomous vehicle sensor system, the capabilities of the motion planning process, and the capabilities of the decision making process, analysis of such connectivity characteristics may allow for a determination that certain road segments or highway gates are potentially not navigable by the autonomous vehicle at a specified level of safety or robustness at certain time(s) or time frame of the day. This may allow the autonomous vehicle to avoid an intersection with, for example, the following geometrical properties: making it impossible for autonomous vehicle sensors to provide data products to the motion planning process in sufficient time to react to impending traffic. In view of the known limitations on the decision making capability of a vehicle, it may also allow an autonomous vehicle to avoid intersections that are too complex to navigate safely (e.g., due to complex required merging, or to be able to deduce travel in a dedicated travel lane).

Spatial orientation of road features

As shown in the example shown in fig. 6, the road network information may contain or allow for calculation of information related to the spatial orientation (e.g., orientation in an inertial coordinate system) of a particular road segment or individual road segment lanes or other road features by a separate process.

In view of the knowledge of the detected attributes of the autonomous vehicle sensor system, analysis of the orientation of road features may allow for a determination that certain road segments or highway gates are potentially not navigable by the autonomous vehicle at a specified level of safety or robustness at a certain time(s) or time frame of the day. This may allow the autonomous vehicle to avoid being "sun blinded" (i.e., experiencing severely degraded performance of the video camera and/or lidar sensor due to exposure to direct sunlight at low oblique incidence angles), for example.

Road construction and location of traffic accident

The road network information may contain or be augmented via a real-time mapping service provider or another input to include information about the location of road construction or accident that potentially caused the closure of certain road segments. In view of the knowledge of the detected properties of the sensor system of the autonomous vehicle, analysis of such information may allow for a determination that certain road segments or highway entrances and exits cannot be navigated by the autonomous vehicle due to the inability of the vehicle to detect temporary signs, obstacles, or gesture signals presented by personnel traffic guidance associated with the road construction or accident.

Location of rough road features

The road network information may contain or be augmented via real-time mapping service provider or similar input to include information regarding the location of rough, degraded, pothole, damaged, wash-board or partially constructed roads, including unprepared roads and secondary roads, and areas deliberately constructed with deceleration strips or vibration strips. This information may be in the form of binary designations (e.g., "rough roads" or "smooth roads") or in the form of continuous numbers or semantic metrics that quantify the roughness of the road surface.

In view of the knowledge of the detected properties of the sensor system of the autonomous vehicle, analysis of road surface roughness may allow for a determination that certain road segments or highway entrances and exits potentially cannot be navigated by the autonomous vehicle at a specified safety or robustness level at a certain time(s) or within a time frame of the day. This may allow the autonomous vehicle to avoid, for example, severely wash-board roads that cause vibrations in the physical sensor mounts resulting in poor sensor system performance, or road areas with speed bumps that may be accidentally classified as non-passable barriers by the perception process.

Position of road feature with poor visibility due to curves and gradients

As shown in fig. 7 and 8, the road network information may contain or allow calculation of information about the curvature and slope (along the vehicle pitch or roll axis) of the road feature by a separate process.

In view of the knowledge of the detected properties of the sensor system of the autonomous vehicle, analysis of the curve and slope of road features may allow for a determination that certain road segments or highway gates are potentially not navigable by the autonomous vehicle at a specified safety or robustness level at certain time(s) or within a time frame of the day. This may allow the autonomous vehicle to avoid road segments as follows: steep and thus makes it impossible for a vehicle sensor system to "look across hills" (i.e., unable to detect the presence of traffic in the surrounding environment due to the limited vertical field of view of the sensor), and "see around corners" (i.e., unable to detect the presence of traffic in the surrounding environment due to the limited horizontal field of view of the sensor).

With illegible, eroded, unintelligible, poorly maintained or positioned marks, signs, or signals Location of road features

The road network information may contain or be augmented via real-time mapping service provider or other input to include information about the location of road areas having: illegible, eroded, unintelligible, or poorly maintained or located lane markings and other roadway markings, signs, or signals.

In view of the knowledge of the detection properties of the sensor system of the autonomous vehicle, and the ability of the (potentially) motion planning or decision making process, analysis of such information may allow for determining that certain road segments or highway gates are potentially not navigable by the autonomous vehicle at a specified level of safety or robustness at a certain time(s) or within a time frame of the day. This may allow autonomous vehicles to avoid, for example, poorly marked road areas to account for the inability of the vehicle to safely navigate within lanes, intersections, etc. having traffic signs or signals as follows: partially obscured (e.g., by foliage) or otherwise difficult to detect from the nominal lane(s). It may also allow autonomous vehicles to avoid road areas with, for example, the following signals or signs: region-specific or country-specific and cannot be reliably detected by the vehicle perception process (es).

Location of road features with poor previous drivability of autonomous or other autonomous vehicles

The road network information may contain, or be augmented via a real-time mapping service provider or another input or by an autonomous vehicle of interest in the autonomous vehicle fleet or any other vehicle to include, information regarding the location of the following road features: in which an autonomous vehicle of interest or another autonomous vehicle has experienced dangerous, degraded, unsafe, or otherwise undesirable drivability, potentially due to high scene traffic or pedestrian density, occlusion of stationary objects, traffic intersection complexity, or other factors. The identification of areas of poor vehicle performance may be "marked" in the map database and marked as avoided when the number of marked events exceeds a specified threshold. This may allow the autonomous vehicle to avoid road features that the vehicle or other vehicle has experienced navigation difficulties.

Location of road features with poor previous simulation performance of model autonomous vehicles

The road network information may contain or be augmented to include information about the location of the road area as follows: in which a model of an autonomous vehicle of interest has been observed in a simulated environment to potentially experience dangerous, degraded, unsafe, or otherwise undesirable drivability due to scene traffic or pedestrian density, occlusion of static objects, complexity of traffic intersections, or other factors. The identification of areas of poor vehicle performance may be "marked" in the map database and marked for avoidance. This may allow the autonomous vehicle to avoid road areas as follows: in which a model of a vehicle has experienced difficulty navigating safely in a simulated environment, indicating that an experimental vehicle may face navigation challenges in a real world environment.

The location of road features that may present physical navigation challenges in severe weather

The road network information may contain, or allow calculation by a separate process, or be augmented via a real-time mapping service provider or another input to include information about the location of road features that may present navigation challenges in severe weather or under specified environmental conditions.

In view of the knowledge of the detection attributes of the sensor system of the autonomous vehicle, as well as the knowledge of the performance characteristics of the motion control process of the vehicle, analysis of such information may allow for a determination that certain road segments or highway gates are potentially not navigable by the autonomous vehicle at a specified level of safety or robustness at a certain time(s) or within a time frame of the day. This may allow the autonomous vehicle to avoid road segments that include, for example, road inclines or curves, which may not be safely navigated when covered by ice, snow, or frost rain.

The location of road features that may lead to known vehicle faults or fault conditions

The road network information may contain, or allow calculation by a separate process, or be augmented via a real-time mapping service provider or another input to include information regarding the location of road features that may lead to known vehicle faults or fault conditions in various sensor or software processes.

In view of the knowledge of the detection attributes of the sensor system of the autonomous vehicle, as well as the knowledge of the performance characteristics of the motion control process of the vehicle, analysis of such information may allow for a determination that certain road segments or highway gates are potentially not navigable by the autonomous vehicle at a specified level of safety or robustness at a certain time(s) or within a time frame of the day. This may allow the autonomous vehicle to avoid certain types of metal bridges or overpasses, for example, that may cause false readings from radar sensors, certain tunnels that may block GPS signals and thus result in poor vehicle positioning performance, and certain extremely flat road areas that may not provide vertical features that may be detected by lidar sensors and may thus result in poor vehicle positioning performance.

It is impossible to safely navigate road segments including road inclination or curve when covered by ice, snow or freezing rain.

The road network information may contain, or allow calculation by a separate process, or be augmented by a real-time mapping service provider or another source to include information about the location of road features that may present navigation challenges in severe weather or under specified environmental conditions.

In view of the knowledge of the detection attributes of the sensor system of the autonomous vehicle, as well as the knowledge of the performance characteristics of the motion control process of the vehicle, analysis of such information may allow for a determination that certain road segments or highway gates are potentially not navigable by the autonomous vehicle at a specified level of safety or robustness at a certain time(s) or within a time frame of the day. This may allow autonomous vehicles to avoid road segments that include, for example, road inclines or curves, which may not be safely navigated when covered by ice or freezing rain.

In addition to identifying a particular road segment that cannot be safely navigated by an autonomous vehicle, it is also possible to perform the opposite operation: based on the analysis of the relevant information sources as described above, a particular road segment that can be safely navigated by the autonomous vehicle is identified. For example, based on analysis of known performance characteristics of vehicle sensors and software processes, as well as given information about road characteristics, it is possible to determine whether a given road segment can be safely and robustly navigated by an autonomous vehicle.

Such analysis would be useful for compiling a feed of map data products or data to be used by other products or processes, describing a "verified autonomous driving route" of the autonomous vehicle. In some implementations, the data product or data feed may describe an "unsafe autonomous driving route". The data may be used as one of the attributes of the road segment that is maintained as part of the road network information. In some cases, verification of road segments and routes (or determination of inability to travel safely or robustly) may be based on successful experimental travel (or simulated travel) of the autonomous vehicle at the level of road features such as streets or at the level of lanes within a given road feature. When determining the optimal route between the current location of the own vehicle and the target location, the routing algorithm may use this information by considering only the verified autonomous driving route. Such an optimal route may attempt to include only road segments that are considered "verified autonomous driving routes," or it may attempt to include a combination of verified and unverified driving routes, where the combination is determined by an optimization process that takes into account various factors, such as distance traveled, expected travel time, and whether the road segments are verified or unverified. In general, the route algorithm may explore only candidate routes known to have a feasibility state exceeding a feasibility threshold, e.g., to allow for sufficiently robust or sufficiently safe travel or sufficiently robust and sufficiently safe travel.

In some instances, such information may be used for city planning purposes to enable a user (i.e., a human planner of a road network or an automated road network planning software process) to avoid designing road segments or intersections that may pose navigation challenges to autonomous vehicles in such use cases, the analysis methods described herein will be used in the context of road design software tools or processes.

Such road design software tools or processes would allow a user to specify or design road segments, road networks, intersections, highways, or other road features using a variety of possible input devices and user interfaces. When a user samples a road design software tool to specify or design road segments, road networks, intersections, roads, or other road features, a software process (i.e., a "feasibility state process") will analyze the feasibility state of road segments or regions in a plurality of potentially connected road segments (e.g., highways (freeway) or intersections). The feasibility state process may also analyze the feasibility state of the route. The feasibility state is determined based on the analysis method described above.

The output of the feasibility state process may be a feasibility state assessment, i.e., an assessment representing the feasibility of a road segment, road network, intersection, road, or other road feature or route in a binary designation (e.g., "feasible" or "infeasible"), or may take the form of a continuous numerical or semantic metric that quantifies the feasibility. The feasibility state assessment may include independent assessments of the safety or robustness of road segments, road networks, intersections, roads, or other road features or routes, each assessment being represented in a binary designation or in the form of a continuous numerical or semantic metric quantifying the safety or robustness. The output of the feasibility state process may include a warning to the user based on the value of the feasibility state evaluation.

Based on the value of the feasibility state evaluation, road segments, road networks, intersections, roads, or other road features designed by the user may be automatically deleted. Depending on the value of the feasibility state assessment, road segments, road networks, intersections, roads, or other road features may be automatically modified in a manner that improves the feasibility state assessment.

In this way, the road design tools or processes may be able to alert a user when the user designs a dangerous road segment, intersection, or route, and thereby prevent construction of such road segment, intersection, or route, and also potentially suggest improved designs for the road segment, intersection, or route

We sometimes use the phrase "feasibility state" to broadly include any determination or indication of a suitability level of a route or road feature or route segment for the travel of an autonomous vehicle, e.g., whether unsafe, safe, robust, and robust, valid, and other similar interpretations.

Other embodiments are within the scope of the following claims.

Claims

1. A computer-implemented method for route planning, comprising:

the computer is used for processing the data of the computer,

determining, by a route planning process, two or more candidate routes, each comprising a sequence of connected road segments, one of the candidate routes for travel by an autonomous vehicle from a starting location to a target location, wherein the one of the candidate routes comprises a road segment that is not included in at least another one of the candidate routes;

predicting, for a route of the two or more candidate routes, a performance level to be achieved by a sensor of the autonomous vehicle if the autonomous vehicle is to travel on the route, wherein the predicted performance level is affected by an environmental condition to which the autonomous vehicle is to be exposed by traveling on the route;

determining a feasibility of the route for which the performance level is predicted based on a predicted performance level to be achieved by the sensor and performance characteristics of at least one of a motion planning process and a decision-making process for each of the two or more candidate routes with respect to the autonomous vehicle; and

The route for which the feasibility is determined to be infeasible is excluded from consideration based on the determined feasibility.

2. The method of claim 1, wherein the feasibility is dependent on characteristics of road features.

3. The method of claim 2, wherein the characteristic of the road feature comprises a spatial characteristic of an intersection, a roundabout, or a highway doorway.

4. The method of claim 2, wherein the characteristic of the road feature comprises a connectivity characteristic of an intersection, a roundabout, or a highway doorway.

5. The method of claim 2, wherein the characteristic of the road feature comprises a spatial orientation.

6. The method of claim 2, wherein the characteristic of the road feature comprises road construction or a traffic accident.

7. The method of claim 2, wherein the characteristic of the road feature comprises roughness.

8. The method of claim 2, wherein the characteristics of the road feature include the following characteristics of visibility due to curvature and slope: indicating that the autonomous vehicle is unable to navigate the road feature at a specified level of safety or a specified level of robustness.

9. The method of claim 2, wherein the characteristic of the road feature comprises the following characteristic of previous drivability of an autonomous vehicle: indicating that the autonomous vehicle is unable to navigate the road feature at a specified level of safety or a specified level of performance.

10. The method of claim 2, wherein the characteristics of the road feature include the following characteristics of a previously simulated performance of a model autonomous vehicle: indicating that the autonomous vehicle is unable to navigate the road feature at a specified level of safety or a specified level of performance.

11. The method of claim 2, wherein the characteristic of the road feature comprises information related to a road feature that presents a navigation challenge in inclement weather, the information indicating that the autonomous vehicle is unable to navigate the road feature at a specified level of safety or a specified level of robustness.

12. A computer-implemented method for route planning, comprising:

the computer is used for processing the data of the computer,

determining, by a route planning process, two or more candidate road segments belonging to a road network information body, the candidate road segments being used to compose a route for travel by an autonomous vehicle from a starting location to a target location;

Predicting, for a road segment of the two or more candidate road segments, a performance level to be achieved by a sensor of the autonomous vehicle if the autonomous vehicle is to travel on the road segment, wherein the predicted performance level is affected by an environmental condition to which the autonomous vehicle is to be exposed by traveling on the road segment;

determining a feasibility of the road segment for which the performance level is predicted based on a predicted performance level to be achieved by the sensor; and

based on the determined feasibility, excluding road segments for which the feasibility is determined to be infeasible from consideration,

wherein the feasibility is further dependent on performance characteristics of a software process and characteristics of the road segments, and the software process includes at least one of a motion planning process and a decision making process for each of the two or more candidate road segments of the autonomous vehicle.

13. The method of claim 12, wherein the characteristic of the road segment comprises a spatial characteristic of an intersection, a roundabout, or a highway doorway.

14. The method of claim 12, wherein the characteristic of the road segment comprises a connectivity characteristic of an intersection, a roundabout, or a highway doorway.

15. The method of claim 12, wherein the characteristic of the road segment comprises a spatial orientation.

16. The method of claim 12, wherein the characteristic of the road segment comprises road construction or a traffic accident.

17. The method of claim 12, wherein the characteristic of the road segment comprises roughness.

18. The method of claim 12, wherein the characteristics of the road segment include the following characteristics of visibility due to curvature and slope: indicating that the autonomous vehicle is unable to navigate the road feature at a specified level of safety or a specified level of robustness.

19. The method of claim 12, wherein the characteristics of the road segment include the following characteristics of previous drivability of an autonomous vehicle: indicating that the autonomous vehicle is unable to navigate the road feature at a specified level of safety or a specified level of performance.

20. The method of claim 12, wherein the characteristics of the road segment include the following characteristics of previously simulated performance of a model autonomous vehicle: indicating that the autonomous vehicle is unable to navigate the road feature at a specified level of safety or a specified level of performance.

21. The method of claim 12, wherein the characteristic of the road segment includes information related to a road feature that presents a navigation challenge in inclement weather, the information indicating that the autonomous vehicle is unable to navigate the road feature at a specified level of safety or a specified level of robustness.