METHOD FOR PLANNING A LANE-LEVEL PATH FOR A VEHICLE IN THE PRESENCE OF SCENE UNCERTAINTY

Abstract

A method for planning a lane-level path for a vehicle comprises generating several possible paths for the vehicle according to a navigation goal, each of which includes a sequence of lane maneuvers for the vehicle. The method further comprises producing several proposed task lists for each of the several possible paths based on at least one uncertainty regarding the sequence of lane maneuvers; determining a utility factor for each of the several proposed task lists; and selecting an optimal path from the several possible paths and an optimal task list from the several proposed task lists, thereby planning the lane-level path for the vehicle to reach the navigation goal.A vehicle contains multiple wheels and a vehicle-side controller in communication with the multiple wheels, which contains a set of commands to execute the procedure.

Description

INTRODUCTION

[0001] The disclosure relates to a method for planning a path at lane level for a vehicle and to a vehicle.

[0002] Vehicles such as automated or autonomous vehicles can use a pre-generated, pre-processed, high-resolution map of a driving scene to plan and execute a lane-level path to a navigation destination. Such high-resolution maps can contain features such as lane geometries, lane types, lane markings, speed limits, and the like. In the absence of such high-resolution maps, a vehicle planning module may encounter uncertainty when generating a navigation plan. SUMMARY

[0003] A method for planning a lane-level path for a vehicle comprises generating multiple possible paths for the vehicle according to a navigation goal, each of which includes a sequence of lane maneuvers for the vehicle. The method further comprises producing multiple proposed task lists for each of the multiple possible paths based on at least one uncertainty regarding the sequence of lane maneuvers. The method also includes determining a utility factor for each of the multiple proposed task lists and selecting an optimal path from the multiple possible paths and an optimal task list from the multiple proposed task lists, thereby planning the lane-level path for the vehicle to reach the navigation goal.

[0004] According to one aspect, generating the multiple possible paths may involve analyzing a driving scene generated by the vehicle, wherein the driving scene contains at least one uncertainty chosen from a vehicle maneuver type, a lane crossing type, and a lane edge type.

[0005] According to an additional aspect, generating the multiple possible paths can involve defining a sequence of one or more lanes for the vehicle, configured to reach the navigation destination.

[0006] According to another aspect, generating the multiple proposed task lists may involve listing several sequential vehicle maneuvers configured to reach each of the multiple possible paths.

[0007] According to another aspect, producing the multiple proposed task lists may involve assigning an occurrence probability to each of the multiple sequential vehicle maneuvers.

[0008] According to one aspect, producing the multiple proposed task lists may involve assigning an overall probability to each of the multiple proposed task lists based on the probability of occurrence of each of the multiple sequential vehicle maneuvers.

[0009] According to an additional aspect, the procedure can also include determining an expected benefit for each of the several possible paths based on the overall probability and the benefit factor.

[0010] According to another aspect, choosing the optimal path can involve selecting one of the several possible paths that has the highest expected benefit, and choosing the optimal task list can involve selecting one of the several proposed task lists that has the highest benefit factor.

[0011] According to another aspect, determining the benefit factor may include taking into account an estimated time to complete each of the several proposed task lists, and / or a complexity of each of the several proposed task lists, and / or an estimated traffic congestion during the completion of each of the several proposed task lists, and / or a risk factor associated with completing each of the several proposed task lists.

[0012] According to one aspect, considering the risk factor may include assessing whether the vehicle crosses a lane boundary while completing each of the several proposed task lists, whether the vehicle disregards a traffic control device while completing each of the several proposed task lists, whether the vehicle disregards a turning lane while completing each of the several proposed task lists, whether the vehicle reaches the end of a lane while completing each of the several proposed task lists, and whether the vehicle disregards a lane marking while completing each of the several proposed task lists.

[0013] In a further embodiment, a method for planning a lane-level path for a vehicle comprises developing an online map from inputs received from multiple vehicle-side sensors, wherein the online map contains a driving scene for the vehicle that has at least one uncertainty selected from a vehicle maneuver type, a lane crossing type, and a lane edge type. The method further comprises generating multiple possible paths for the vehicle according to a navigation destination, each of the multiple possible paths comprising a sequence of lane maneuvers for the vehicle. Additionally, the method comprises producing multiple proposed task lists for each of the multiple possible paths based on the at least one uncertainty and determining a utility factor for each of the multiple proposed task lists.The procedure also includes selecting an optimal path from the multiple possible paths and an optimal task list from the multiple suggested task lists, in order to plan the path at lane level for the vehicle to reach the navigation destination.

[0014] According to one aspect, developing the online map can include creating a scene graph from the driving scene.

[0015] According to an additional aspect, generating the multiple possible paths can involve searching the scene graph using a depth-first search algorithm.

[0016] According to another aspect, producing the multiple proposed task lists may involve a step-by-step evaluation of each of the multiple possible paths and a listing of multiple sequential vehicle maneuvers configured to achieve each of the multiple possible paths.

[0017] According to another aspect, producing the multiple suggested task lists can include listing both an actionable suggested task list that enables the navigation goal to be reached, and an unactionable suggested task list that does not enable the navigation goal to be reached.

[0018] According to one aspect, determining the benefit factor may involve considering a risk factor associated with completing each of the several proposed task lists, the risk factor including whether the vehicle crosses a lane boundary, disregards a traffic control device, disregards a turning lane, reaches the end of a lane, and disregards a lane marking while completing each of the several proposed task lists.

[0019] According to an additional aspect, the procedure can also include automated control of the vehicle along the path at lane level.

[0020] A vehicle contains multiple wheels configured to be offset along a path at lane level, and a propulsion power source configured to drive the multiple wheels.The vehicle also includes a vehicle-side controller communicating with the propulsion power source, which contains a set of instructions executable to: generate multiple possible paths for the vehicle according to a navigation destination, each of the multiple possible paths comprising a sequence of lane maneuvers for the vehicle; produce multiple proposed task lists for each of the multiple possible paths based on at least one uncertainty regarding the sequence of lane maneuvers; determine a utility factor for each of the multiple proposed task lists; and select an optimal path from the multiple possible paths and an optimal task list from the multiple proposed task lists, thereby planning the lane-level path for the vehicle to reach the navigation destination.

[0021] According to one aspect, the command set may include a planning module configured to plan the path at lane level in order to reach the navigation destination in the presence of at least one uncertainty without access to a high-resolution map of the multiple possible paths.

[0022] According to an additional aspect, the vehicle-side controller can be configured to automatically control the drive power source.

[0023] The features and advantages described above, as well as further features and accompanying advantages of this disclosure, will become apparent from the following detailed description of illustrative examples and modes of carrying out the present disclosure when considered in conjunction with the accompanying drawings and the attached claims. Furthermore, this disclosure expressly includes combinations and subcombinations of the elements and features illustrated above and below. BRIEF DESCRIPTION OF THE DRAWINGS Fig. Figure 1 is a schematic representation of a side view of a vehicle having a vehicle-side controller containing a set of instructions for executing a procedure for planning a lane-level path for a vehicle. Fig. Figure 2 is a schematic representation of the procedure performed by the vehicle-side controller of Fig. 1 is executable. Fig. Figure 3 is a schematic representation of a further embodiment of the method of Fig. 2. Fig. 4 is a schematic representation of a top view of an online map containing a driving scene, which is traversed by the vehicle of Fig. 1 is generated. Fig. 5 is a schematic representation of a scene graph derived from the driving scene of Fig. 4 is created. DETAILED DESCRIPTION

[0024] Referring to the figures, where similar reference signs refer to similar elements, a vehicle 10 ( Fig. 1) and a procedure 12, 112 ( Fig. 2 and Fig. 3) to plan a path 14 at lane level ( Fig. 5) for vehicle 10 in general. Method 12, 112 may be useful for applications requiring lane-level path planning for vehicles 10 without access to a high-resolution map of a driving scene 16 ( Fig. 4) for the vehicle 10. For example, a high-resolution map may contain specific, precise details regarding painted lane lines or markings; three-dimensional building models; traffic signs; intersection control types; traffic lights; stop lines; lane types; lane junctions; roadway types for densely populated urban areas, commercial and industrial zones, and suburban and residential areas; and fixed physical installations along highways, parking lots, garages, driveways, alleys, and the like. The absence of one or more of these map features may introduce uncertainty into the path planning for a vehicle 10. However, Procedure 12, 112, and the vehicle 10 can eliminate dependence on such high-resolution maps for path planning.

[0025] In particular, the method 12, 112 can be useful for planning the path 14 at lane level, even if there is uncertainty in the driving scene 16, e.g., regarding lane geometry, topology, and features, as explained in more detail below. That is, the method 12, 112 can be used to create an online map 46 ( Fig. 4), for example, to provide lane geometries, lane features, and connections between lanes, and to plan the path 14 at lane level when there is an uncertain environment around the vehicle 10. Therefore, and as will be explained in more detail below, the method 12, 112 can enable automated or autonomous driving in the absence of a high-resolution map and in the presence of scene uncertainty.

[0026] Therefore, the method 12, 112 and the vehicle 10 can be useful for motor vehicle applications such as, but not limited to, internal combustion engine vehicles, electric vehicles, hybrid vehicles and the like. For example, the vehicle 10 can be a motor vehicle powered by a drive power source 18 ( Fig. 1) is supplied with energy that powers an internal combustion engine 20 ( Fig. 1) and / or an electric motor 22 ( Fig. 1) and / or contains an energy storage device 24.

[0027] Furthermore, the vehicle 10 can be configured for autonomous or automated driving, whereby the vehicle 10 can be controlled or driven by technology comprising hardware and software located remotely from or on board the vehicle 10, enabling the vehicle 10 to drive without active physical control by a human operator. For example, autonomous or automated driving tasks may include, but are not limited to, the detection, recognition, and classification of objects and events; a response to objects and events; maneuver planning; steering, turning, lane keeping, signaling, and lane changes; and acceleration and deceleration.

[0028] Alternatively, Method 12, 112 and Vehicle 10 may be useful for non-motor vehicle applications, such as, but not limited to, space, aviation, marine, public transportation, agricultural, industrial, and rail applications. For example, Vehicle 10 may be, but is not limited to, a commercial vehicle, an industrial vehicle, a passenger car, an aircraft, a watercraft, a train, a wagon, a bus, or the like. It is also envisaged that, to achieve the purposes of this disclosure, Vehicle 10 may be a mobile platform such as an airplane, a quad bike (ATV), a boat, a personal mobility device, a robot, and the like.

[0029] Now, with reference to Fig. 1 the vehicle 10 several wheels 26, which are configured along the path 14 at lane level ( Fig. 5) to be displaced. That is, the multiple wheels 26 can be steerable and / or non-steerable wheels and can be configured to travel or walk along a ground surface of the path 14 at lane level.

[0030] With renewed reference to Fig. 1 The vehicle 10 also includes the drive power source 18, which is configured to drive the multiple wheels 26. In certain embodiments, the drive power source 18 may include an energy storage device 24, which is configured to store and release electrical energy. For example, the energy storage device 24 may be a rechargeable high-voltage battery or battery assembly. As in Fig. As shown in Figure 1, the vehicle 10 can include several energy storage devices 24 that are electrically connected to each other to provide an output power to the vehicle 10. Furthermore, the energy storage device 24 can be configured to provide a driving force to at least one of the several wheels 26. That is, in certain embodiments, the vehicle 10 can be an electric vehicle 10 that receives a driving force from the energy storage device 24. Therefore, the vehicle 10 can include one or more electric motors 22, each assigned to one of the several wheels 26, and be configured to drive the several wheels 26.

[0031] In further embodiments, the drive power source 18 can include an internal combustion engine 20, which can drive the multiple wheels 26 by means of a transmission (not shown), a drive shaft (not shown), a differential (not shown), and / or an axle (not shown). Additionally or alternatively, the internal combustion engine 20 can interact with the energy storage device 24 to provide drive power to the multiple wheels 26.

[0032] With renewed reference to Fig. 1 The vehicle 10 also contains a vehicle-side controller 28 in communication with the propulsion power source 18, which contains a set of commands that can be executed to control the path 14 at lane level ( Fig. 5) to plan for the vehicle 10, as detailed below. That is, the controller 28 can be on board the vehicle 10 and can execute the procedure 12, 112 for planning the path 14 at lane level, as described below. In particular, the vehicle-side controller 28 can include a processor configured to execute programmed code and can run an operating system. The processor can include read / write memory (RAM) and a storage device such as a hard disk. The vehicle-side controller 28 can include programming to analyze data from the vehicle 10 and diagnose the existence of a precondition of the procedure 12, 112.

[0033] The vehicle-side controller 28 can be used for automated or autonomous control 50 ( Fig. 3) the drive power source 18 and the vehicle 10 are configured. That is, the vehicle-side controller 28 can contain programming to perform actions such as detecting, recognizing, and classifying objects and events; reacting to objects and events; planning maneuvers; steering, turning, keeping in lane, signaling, changing lanes, and accelerating and decelerating the vehicle 10 without active physical control by a human operator.

[0034] The vehicle-side controller 28 can also contain programming to further perform actions regarding aspects of the procedure 12, 112 such as generating 30 ( Fig. 2 and Fig. 3) several possible paths 32 ( Fig. 5) for vehicle 10; produce 34 ( Fig. 2 and Fig. 3) several proposed task lists for each of the several possible paths 32; Determine 36 ( Fig. 2 and Fig. 3) a utility factor (u) for each of the several proposed task lists; Choose 38 ( Fig. 2 and Fig. 3) of an optimal path and an optimal task list in order to plan the path 14 at lane level for the vehicle 10; completion of the procedure 12, 112; electrical communication with the drive power source 18 or other components of the vehicle 10; monitoring the driving behavior of the vehicle 10 and the like.

[0035] In particular, it contains, as below with reference to Fig. 2 and Fig. 3 is explained and described in more detail, the vehicle-side controller 28 the instruction set that is executable to generate 30 of the several possible paths 32 ( Fig. 5) for vehicle 10 according to a navigation destination 40 ( Fig. 4); Produce 34 of the several proposed task lists for each of the several possible paths 32 based on at least one uncertainty; Determine 36 the utility factor (u) for each of the several proposed task lists and select 38 the optimal path and the optimal task list to thereby make the path 14 at lane level ( Fig. 5) to plan for vehicle 10 to reach navigation destination 40. In particular, as explained in more detail below, the command set may include a planning module configured to plan path 14 at lane level to reach navigation destination 40 in the presence of at least one uncertainty without access to a high-resolution map of the multiple possible paths 32.

[0036] Additionally, although not shown exactly, the vehicle 10 may contain a communication bus configured to enable electronic communication between components of the vehicle 10. The propulsion power source 18 and the vehicle 10 may contain sensors, and the sensors, the vehicle-side controller 28, and the propulsion power source 18 may be electrically connected to the communication bus and may transmit data and computer-aided instructions via it to carry out the aspects of procedure 12, 112. Generation of multiple possible paths

[0037] With renewed reference to Fig. 2 includes the procedure 12 for planning the path 14 at lane level for the vehicle 10, a generation 30 of the several possible paths 32 ( Fig. 5) for vehicle 10 according to navigation destination 40 ( Fig. 4) The vehicle 10 or the planning module can receive the navigation destination 40 as an input from a user or from a navigation system of the vehicle 10. As referred to in Fig. As described in section 4, a non-restrictive example of the navigation goal 40 for the vehicle 10 could be to turn right at an intersection 42 in 150 meters. The procedure 12 then includes developing or generating 30 each of the several possible paths 32 or plans for the vehicle 10 to or for achieving the navigation goal 40.

[0038] In particular, as with reference to Fig. 4 and Fig. As described in section 5, generating 30 of the several possible paths 32 involves analyzing a driving scene 16 ( Fig. 4), which is generated by the vehicle 10. That is, while driving, the vehicle 10 can scan or monitor the driving scene 16 with several vehicle-side sensors, such as passive or active sensors including cameras, LiDAR, radar, and the like. Because the driving scene 16 can be analyzed on the fly while the vehicle 10 is driving, the driving scene 16 may have less detail and precision than a high-resolution map and may therefore contain at least one uncertainty.

[0039] In the form of non-restrictive examples, driving scene 16 can contain at least one uncertainty chosen from a vehicle maneuver type, a lane intersection type, and a lane edge type. That is, although further uncertainties may exist, driving scene 16 can contain uncertainty or lack of confidence, for example, regarding which lane to traverse, whether a solid or dashed lane marking exists between two adjacent lanes, whether a lane exists, whether a lane permits turning, whether intersection 42 is controlled by a traffic light or a traffic sign, whether intersection 42 is uncontrolled, and the like.

[0040] As with continued reference to Fig. As described in Figure 4, several possible lanes are designated as Lane 1 (L1), Lane 2 (L2), Lane 3 (L3), and so on. When vehicle 10 is traveling in Lane 1 (L1), as shown, there may be uncertainty or lack of confidence regarding the next maneuver permitted for vehicle 10. That is, as shown, vehicle 10 may be permitted or allowed to continue straight ahead in Lane 1 (L1), or it may be permitted to continue straight ahead and turn right using Lane 8 (L8) to proceed to Lane 10 (L10). Similarly, when vehicle 10 is traveling in Lane 4 (L4), there may be uncertainty regarding the next permitted maneuver for vehicle 10.This means the vehicle may be permitted to continue straight ahead in lane 4 (L4), it may be permitted to continue straight ahead using lane 7 (L7) and turn right to proceed to lane 9 (L9), or it may be permitted to turn right.

[0041] Similarly, there may be uncertainty regarding the intersection type for intersection 42, which is located at lanes 2 (L2), 5 (L5), 7 (L7), and 8 (L8), i.e., intersection 42 is located before lanes 3 (L3) and 6 (L6). For example, intersection 42 may be controlled by traffic lights, may be uncontrolled and include a stop sign, or may be uncontrolled and include a yield sign.

[0042] Furthermore, there may be uncertainty regarding the lane edge type for adjacent lanes 1 (L1) and 4 (L4). For example, depending on local traffic laws and lane marking specifications, lanes 1 (L1) and 4 (L4) may be separated by a dashed line, a single solid line, or a double solid line.

[0043] Therefore, in order to take such uncertainties into account and as with continued reference to Fig. As described in section 4, the generation of 30 ( Fig. 2) The multiple possible paths 32 for vehicle 10 include a specification of a sequence of one or more lanes for vehicle 10 configured to reach navigation destination 40. That is, each of the multiple possible paths 32 contains a sequence of lane maneuvers for vehicle 10. In other words, the multiple possible paths 32 can be lane-level paths or plans to reach navigation destination 40.

[0044] In one embodiment, which refers to Fig. As described in section 3, the procedure 112 can be used to develop 44 an online map 46 ( Fig. 4) from inputs received from the multiple vehicle sensors. That is, the online map 46 can, as in Fig. Figure 4 shows the available possible paths 32 and the connections between the multiple possible paths 32. In other words, the online map 46 can provide the lanes, lane connections, and features for planning the path 14 at lane level and can provide a starting point for determining the possible routes to reach the navigation destination 40, given the driving scene 16 and the inherent uncertainty.

[0045] Furthermore, developing 44 the online map 46 can create a scene graph 48 ( Fig. 5) from driving scene 16 ( Fig. 4) included. That is, for the scene graph 48, each lane can be represented as a node, and each lane maneuver can be represented as a connection or a branch between nodes. Since the online map 46 can also be developed during operation while the vehicle 10 is driving, the online map 46 may have less detail and precision than a high-resolution map and may therefore contain at least one uncertainty. For example, the online map 46 may also contain at least one uncertainty regarding the vehicle maneuver type, the lane intersection type, and the lane edge type.

[0046] In a non-restrictive example, generating 30 of the multiple possible paths 32 can involve searching the scene graph 48 using a depth-first search algorithm. That is, each of the multiple possible paths 32 can be generated using a depth-first search algorithm. For example, the depth-first search algorithm can be a recursive algorithm that traverses or searches the scene graph 48 by starting at a root node and examining each branch as far as possible before tracing back, in order to generate the multiple possible paths 32 30 that contain the sequence of lane maneuvers according to the navigation goal 40. For example, so that vehicle 10, which is in lane 1 (L1) of Fig. 4, the navigation destination 40 of the right turn is reached in 150 meters, the several possible paths 32 or plans include the following: (I) Possible path or plan 1: L1 → L8 → L10 Possible path or plan 2: L1 + L4 → L7 → L9 Producing the several proposed task lists

[0047] With renewed reference to Fig. 2. Procedure 12 further includes producing 34 of the multiple proposed task lists for each of the multiple possible paths 32 based on at least one uncertainty regarding the sequence of lane maneuvers. That is, producing 34 of the multiple proposed task lists may include listing multiple sequential vehicle maneuvers configured to reach each of the multiple possible paths 32. In particular, producing 34 of the multiple proposed task lists may include stepwise evaluating each of the multiple possible paths 32 and listing the multiple sequential vehicle maneuvers configured to reach each of the multiple possible paths 32. For example, a new scene graph 48 may be generated and traversed to generate the multiple proposed task lists for each of the multiple possible paths 32, as detailed below.

[0048] Furthermore, producing 34 of the multiple proposed task lists can include listing both a feasible proposed task list that enables reaching navigation destination 40 and a non-feasible proposed task list that does not enable reaching navigation destination 40. For example, a feasible proposed task list can enable vehicle 10 to proceed to navigation destination 40. In contrast, a non-feasible proposed task list can result in a dead end for vehicle 10.

[0049] Therefore, the multiple proposed task lists can be multiple hypotheses containing specific tasks or maneuvers for vehicle 10. As a non-restrictive example, procedure 12, for the multiple possible paths 32 outlined above in (I), can produce 34 of the following proposed task lists for each of the multiple possible paths 32 for vehicle 10 positioned in lane 1 (L1). Fig. 4 for navigation destination 40, turning right at intersection 42 in 150 meters, includes the following: (II) Possible path or plan 1: L1 → L8 → L10 Task list or hypothesis 1 - Feasible 1. Follow the lane 2. Stopping at a traffic light 3. Turn right at the intersection Task list or Hypothesis 2 - dead end or not feasible 1. Follow the lane 2. Stopping at a traffic light 3. No connection on the right Possible path or plan 2: L1 → L4 → L7 → L9 Task list or hypothesis 1 - Feasible 1. Change lanes to the right 2. Stopping at a traffic light 3. Turn right at the intersection Task list or Hypothesis 2 - dead end or not feasible 1. Follow the lane 2. Lane change not possible Task list or Hypothesis 3 - Feasible 1. Change lanes to the left 2. Stopping at a stop sign 3. Turn right at the intersection Task list or Hypothesis 4 - dead end or not feasible 1. Change lanes to the left 2. Stopping at a traffic light 3. No connection on the right Probability of occurrence

[0050] Additionally, producing 34 of the multiple proposed task lists can involve assigning an occurrence probability (p) to each of the multiple sequential vehicle maneuvers. That is, each of the multiple sequential vehicle maneuvers in each task list need not have the same occurrence probability. For example, it may be more likely that vehicle 10, which is in lane 1 (L1) of Fig. The probability of changing lanes (p) before turning right is higher than remaining in lane 1 (L1). This means that changing lanes from lane 1 (L1) to lane 4 (L4) has a higher probability of occurrence (p) than remaining in lane 1 (L1).

[0051] That is, the assigned probability (p) refers to a confidence value for a corresponding element of the scene graph 48. For example, as with continued reference to Fig. As described in paragraph 4, a vehicle maneuver such as "turn right at intersection" may have a probability (p) of 0.2 because the turning lane 8 (L8) in scene graph 48 may have an existence confidence, i.e., a confidence that turning lane 8 (L8) actually exists, of 0.2. In other words, the probability (p) can be derived from scene uncertainties of scene graph 48. Therefore, a non-restrictive example of assigned occurrence probabilities (p) for the task lists set out in (II) may include the following: (III) Possible path or plan 1: L1 → L8 → L10 Task list or hypothesis 1 - Feasible 1. Follow the lane (p = 1) 2. Stopping at a traffic light (p = 1) 3. Turn right at the intersection (p = 0.2) Task list or Hypothesis 2 - dead end or not feasible 1. Follow the lane (p = 1) 2. Stopping at a traffic light (p = 1) 3. No connection on the right (p = 0.8) Possible path or plan 2: L1 → L4 → L7 → L9 Task list or hypothesis 1 - Feasible 1. Lane change to the right (p = 0.8) 2. Stopping at a traffic light (p = 0.9) 3. Turn right at the intersection (p = 0.9) Task list or Hypothesis 2 - dead end or not feasible 1. Follow the lane (p = 0.2) 2. Lane change not possible Task list or Hypothesis 3 - Feasible 1. Lane change left (p = 0.8) 2. Stopping at a stop sign (p = 0.1) 3. Turn right at the intersection (p = 0.9) Task list or Hypothesis 4 - dead end or not feasible 1. Lane change left (p = 0.8) 2. Stopping at traffic lights (p = 0.9) 3. No connection on the right (p = 0,1) Overall probability

[0052] Furthermore, producing 34 of the multiple proposed task lists may involve assigning an overall probability (P) to each of the multiple proposed task lists based on the probability of occurrence (p) of each of the multiple sequential vehicle maneuvers. That is, the overall probability (P) can be calculated by multiplying together the probability of occurrence (p) for each respective sequential vehicle maneuver. For example, an overall probability (P) for task list 1 of possible path or plan 1 may be equal to p1 × p2 × p3 = 1 × 1 × 0.2 = 0.2. Therefore, a non-restrictive example of assigned overall probabilities (P) for each of the proposed task lists set out in (III) may include the following: (IV) Possible path or plan 1: L1 → L8 → L10 Task list or hypothesis 1 (P = 0.2) - Feasible 1. Follow the lane (p = 1) 2. Stopping at a traffic light (p = 1) 3. Turn right at the intersection (p = 0.2) Task list or hypothesis 2 (P = 0.8) - dead end or not feasible 1. Follow the lane (p = 1) 2. Stopping at a traffic light (p = 1) 3. No connection on the right (p = 0.8) Possible path or plan 2: L1 → L4 → L7 → L9 Task list or hypothesis 1 (P = 0.65) - Feasible 1. Lane change to the right (p = 0.8) 2. Stopping at a traffic light (p = 0.9) 3. Turn right at the intersection (p = 0.9) Task list or hypothesis 2 (P = 0.2) - dead end or not feasible 1. Follow the lane (p = 0.2) 2. Lane change not possible Task list or hypothesis 3 (P = 0.07) - Feasible 1. Lane change left (p = 0.8) 2. Stopping at a stop sign (p = 0.1) 3. Turn right at the intersection (p = 0.9) Task list or hypothesis 4 (P = 0.07) - dead end or not feasible 1. Lane change left (p = 0.8) 2. Stopping at traffic lights (p = 0.9) 3. No connection on the right (p = 0,1) Determining the utility factor

[0053] With renewed reference to Fig. 2. Procedure 12 further includes determining 36 the utility factor (u) for each of the several proposed task lists. The utility factor (u) may be a reward factor or the opposite of a cost factor and may assist in weighting the favorableness of each of the several proposed task lists. A comparatively higher utility factor (u) may indicate a better chance that the vehicle 10 will reach a target lane or perform a target maneuver. For example, determining 36 the utility factor (u) may involve considering an estimated time to complete each of the several proposed task lists; and / or a complexity of each of the several proposed task lists; and / or an estimated traffic congestion during the completion of each of the several proposed task lists; and / or a risk factor associated with completing each of the several proposed task lists.

[0054] For example, considering the estimated time to complete each of the several proposed task lists might involve calculating an estimated time to complete each task list to provide a time component, normalizing the time component, and multiplying the time component by a time coefficient. Similarly, considering the complexity of each of the several proposed task lists might involve calculating a maneuver complexity for each task list to provide a complexity component, normalizing the complexity component, and multiplying the complexity component by a complexity coefficient.Similarly, taking the estimated traffic congestion into account while completing each of the several proposed task lists may involve calculating a traffic congestion estimate for each task list, providing a congestion component, normalizing the congestion component, and multiplying the congestion component by a congestion coefficient.

[0055] Furthermore, considering the risk factor associated with completing each of the several proposed task lists may, in the form of non-restrictive examples, include assessing whether the vehicle 10 crosses a lane boundary while completing each of the several proposed task lists; whether the vehicle 10 disregards a traffic control device while completing each of the several proposed task lists; whether the vehicle 10 disregards a turning lane while completing each of the several proposed task lists; whether the vehicle 10 reaches the end of a lane while completing each of the several proposed task lists; and whether the vehicle 10 disregards a lane marking while completing each of the several proposed task lists.This means that taking the risk factor into account can help determine the suitability of each of the several proposed task lists.

[0056] In a non-restrictive example, for a task list that includes changing lanes across a dashed lane marking, the risk factor might be low, and the benefit factor (u) might be assigned a positive value, e.g., 5. However, for a task list that includes crossing a solid lane marking, the risk factor might be increased to medium, and the benefit factor (u) might be assigned a negative value, e.g., -2, to penalize the task list. Alternatively, for a task list that includes changing lanes across a double solid lane marking, where lane changes are not permitted, the risk factor might be increased even further, and the benefit factor (u) might be assigned a comparatively higher negative value, e.g., -3, to further increase the penalty associated with the task list.

[0057] Therefore, as a non-restrictive example, any task list set out in (IV) can contain the utility factor (u), e.g.: (V) Possible path or plan 1: L1 → L8 → L10 Task list or hypothesis 1 (P = 0,2; u = 3) - Feasible 1. Follow the lane (p = 1) 2. Stopping at a traffic light (p = 1) 3. Turn right at the intersection (p = 0.2) Task list or hypothesis 2 (P = 0.8; u = -2) - dead end or not feasible 1. Follow the lane (p = 1) 2. Stopping at a traffic light (p = 1) 3. No connection on the right (p = 0.8) Possible path or plan 2: L1 → L4 → L7 → L9 Task list or hypothesis 1 (P = 0.65; u = 6) - Feasible 1. Lane change to the right (p = 0.8) 2. Stopping at a traffic light (p = 0.9) 3. Turn right at the intersection (p = 0.9) Task list or hypothesis 2 (P = 0.2; u = -2) - dead end or not feasible 1. Follow the lane (p = 0.2) 2. Lane change not possible Task list or hypothesis 3 (P = 0.07; u = 4) - Feasible 1. Lane change left (p = 0.8) 2. Stopping at a stop sign (p = 0.1) 3. Turn right at the intersection (p = 0.9) Task list or hypothesis 4 (P = 0.07; u = -1) - dead end or not feasible 1. Lane change left (p = 0.8) 2. Stopping at traffic lights (p = 0.9) 3. No connection on the right (p = 0,1) Expected benefits

[0058] With renewed reference to Fig. 2. Procedure 12 can further include determining 136 an expected utility (U) for each of the several possible paths 32 based on the total probability (P) for each task list and the utility factor (u) for each task list. That is, the expected utility (U) can be the sum of the products of the total probability (P) of each task list and the utility (u) of each task list. For the example set out in (V) above, the expected utility (U) for each of the several possible paths 32 can be calculated as follows: (VI) Possible path or plan 1: U = ∑P i u i = [(0.2 × 3) → (0.8 × -2)] = -1 Possible path or plan 2: U = ∑P i u i = [(0.65 × 6) → (0.2 × -2) → (0.07 × 4) → (0.07 × -1)] = 1.45 Selecting the optimal path and task list

[0059] With renewed reference to Fig. In section 2, procedure 12 also includes selecting 38 an optimal path from the several possible paths 32 and an optimal task list from the several proposed task lists, in order to plan path 14 at lane level for vehicle 10 to reach navigation destination 40. That is, selecting 38 the optimal path may involve choosing one of the several possible paths 32 that has the highest expected utility (U), and selecting 38 the optimal task list may involve choosing one of the several proposed task lists that has the highest utility factor (u). For example, the von Neumann-Morgenstern utility theorem can be used to select the optimal possible path 32 and the proposed task list. Therefore, for the non-restrictive example set out above in (VI), possible path or possible plan 2 has the highest expected utility (U), i.e.,1.45 > -1, and task list or hypothesis 1 has the highest utility factor (u), i.e., 6 > -2, 6 > 4, and 6 > -1. Therefore, in this example, procedure 12 can involve selecting 38 a possible path 2 and a possible task list 1 in order to plan path 14 at lane level for vehicle 10 to reach navigation destination 40.

[0060] Additionally, as with reference to Fig.As described in section 3, the method 112 further comprises automated or autonomous control 50 of the vehicle 10 along path 14 at lane level. That is, the vehicle-side controller 28 can receive and execute commands to allow the vehicle 10 to automatically traverse path 14 at lane level, e.g., to control the vehicle 10 by changing lanes to the right, stopping at the traffic light, and turning right at the intersection 42, to move from lane 1 (L1) to lane 4 (L4) to lane 7 (L7) to lane 9 (L9) in order to reach the navigation destination 40 even in the presence of uncertainty in the driving scene 16.

[0061] Therefore, in summary, method 12, 112 can be a stochastic utility-based planning solution attempt to generate an optimal path at lane level 32 and tasks to reach the navigation destination 40, while taking into account uncertainty in the driving scene 16 and occurrence probabilities (p). Thus, method 12, 112 and vehicle 10 can enable automated or autonomous driving without recourse to high-resolution maps. That is, method 12, 112 can advantageously and specifically plan the path 14 at lane level for vehicle 10 even in the presence of uncertainty in the driving scene 16 and even in the absence of a high-resolution map of the driving scene 16. Furthermore, method 12, 112 does not have to address any explainability or behavioral problems, e.g.,how and why the procedure 12, 112 develops the path 14 at lane level, but can instead provide a methodical logical path 14 at lane level for the vehicle 10 based on the driving scene 16.

[0062] The embodiments described in this disclosure are intended to serve as non-limiting examples, and further embodiments may take various and alternative forms. Additionally, the accompanying drawings are not necessarily to scale and may represent a somewhat simplified depiction of various features of this disclosure, including, for example, specific dimensions, orientations, locations, and shapes. Details associated with such features are partly determined by the intended application and the environment in which the described embodiments are used.

[0063] For the purposes of this description, and unless specifically excluded, the use of the singular includes the plural and vice versa; the terms "and" and "or" are to be understood as both connecting and separating; and the words "contain," "include," "comprise," "exhibit," and the like are to mean "contain without limitation." Furthermore, words of approximation such as "about," "essentially," "generally," "approximately," etc., may be used here to mean "at, near, or close to" or "within 0-5% of" or "within acceptable manufacturing tolerances," or logical combinations thereof. As used here, a component that is "configured" to perform a specified function can perform the specified function without modification, rather than merely having the potential to perform the specified function after further modification.In other words, if the described hardware is explicitly configured to perform the specified function, it is specifically chosen, created, implemented, deployed, programmed, and / or designed for the purpose of performing that specified function. Additionally, the use of ordinal numbers such as first, second, and third does not necessarily imply any order, but can instead simply distinguish between multiple instances of a process or structure.

[0064] The detailed description and the drawings or figures are supporting and descriptive for the present instructions; however, the scope of the present instructions is defined exclusively by the claims. While some of the best methods and further embodiments for carrying out the present instructions have been described in detail, various alternative designs and embodiments for practicing the present instructions, as defined in the appended claims, are present. Furthermore, this disclosure expressly includes combinations and subcombinations of the elements and features illustrated above and below.

Claims

[1] Method for planning a path at lane level for a vehicle, the method comprising: Generating multiple possible paths for the vehicle according to a navigation destination, each of the multiple possible paths comprising a sequence of lane maneuvers for the vehicle; Producing multiple proposed task lists for each of the multiple possible paths based on at least one uncertainty regarding the sequence of lane maneuvers; Determining a utility factor for each of the several proposed task lists and Selecting an optimal path from the multiple possible paths and an optimal task list from the multiple suggested task lists, in order to plan the path at lane level for the vehicle to reach the navigation destination. [2] Method according to claim 1, wherein generating the multiple possible paths comprises analyzing a driving scene generated by the vehicle and the driving scene contains at least one uncertainty selected from a vehicle maneuver type, a lane crossing type and a lane edge type. [3] Method according to claim 1, wherein generating the multiple possible paths comprises defining a sequence of one or more lanes for the vehicle configured to reach the navigation destination. [4] Method according to claim 1, wherein producing the multiple proposed task lists comprises listing multiple sequential vehicle maneuvers configured to achieve each of the multiple possible paths. [5] Method according to claim 4, wherein producing the multiple proposed task lists comprises assigning an occurrence probability to each of the multiple sequential vehicle maneuvers. [6] Method according to claim 5, wherein producing the multiple proposed task lists comprises assigning an overall probability to each of the multiple proposed task lists based on the probability of occurrence of each of the multiple sequential vehicle maneuvers. [7] The method of claim 6, further comprising determining an expected benefit for each of the several possible paths based on the overall probability and the benefit factor. [8] The method of claim 7, wherein selecting the optimal path comprises selecting one of the several possible paths which has the highest expected benefit, and Choosing the optimal task list involves selecting one of the several proposed task lists that has the highest utility factor. [9] Method according to claim 1, wherein determining the benefit factor includes taking into account an estimated time to complete each of the multiple proposed task lists, and / or a complexity of each of the multiple proposed task lists, and / or an estimated traffic congestion during the completion of each of the multiple proposed task lists, and / or a risk factor associated with completing each of the multiple proposed task lists. [10] Method according to claim 9, wherein taking the risk factor into account includes assessing whether the vehicle crosses a lane boundary while completing each of the several proposed task lists, whether the vehicle disregards a traffic control device while completing each of the several proposed task lists, whether the vehicle disregards a turning lane while completing each of the several proposed task lists, whether the vehicle reaches the end of a lane while completing each of the several proposed task lists, and whether the vehicle disregards a lane marking while completing each of the several proposed task lists.

Images