Vehicle and safe parking control unit and control method
By introducing control units and machine learning models into vehicles, the problem of safe parking in scenarios with multiple lane changes has been solved, achieving safe and effective parking control that adapts to complex traffic environments.
Patent Information
- Application Number
- CN202211033195.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-08-26
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2042-08-26
AI Technical Summary
Existing technologies lack effective solutions to safely stop vehicles in multiple lane-change scenarios, especially when vehicle software or hardware fails or the driver loses control, making it impossible to reasonably allocate braking distance and lane-change speed to ensure safe stopping.
A control unit is employed, comprising a determination module, a creation module, a calculation module, an evaluation module, and an allocation module. By determining the target parking position, a solution space for the combination of braking parameters is created, a braking parameter allocation tree is calculated, lane change safety is evaluated, and braking distance and lane change speed are allocated. Supervised learning and reinforcement learning models are used to predict environmental information to optimize the parking process.
It achieves reasonable allocation of safe parking in multiple lane change scenarios, improves the safety and efficiency of vehicle parking process, adapts to dynamic environmental changes, reduces the training difficulty of machine learning models and improves generalization ability.
Smart Images

Figure CN117657118B_ABST
Abstract
Description
Technical Field
[0001] This invention relates generally to vehicle driving assistance technology, and more particularly to a safe parking control unit and control method suitable for multiple lane change scenarios. It also relates to a vehicle equipped with the safe parking control unit and a corresponding machine-readable storage medium. Background Technology
[0002] Vehicle driver assistance is a current research hotspot, with automatic lane changing and safe parking being crucial technologies. For example, in emergency situations such as software or hardware malfunctions or driver illness rendering the vehicle uncontrollable, it's essential to stop the vehicle safely and immediately. Existing solutions can safely stop the vehicle within its lane or, after a single lane change, in an adjacent lane. However, for scenarios requiring multiple lane changes before stopping, current technologies lack effective solutions. Summary of the Invention
[0003] Against this backdrop, the present invention proposes a safe parking solution applicable to multiple lane change scenarios, which can reasonably allocate the braking distance and lane change speed of the vehicle in each lane it passes through during the parking process.
[0004] According to one aspect of the present invention, a control unit for safe parking of a vehicle is provided, comprising: a determination module configured to determine a target parking position of the vehicle on a multi-lane road, wherein the distance between the current position of the vehicle and the target parking position includes multiple braking distances on multiple lanes traversed by the vehicle during the parking process; a creation module configured to create a solution space for braking parameters of each of the multiple lanes, including multiple candidate braking distances on each lane, and a lane change speed and lane change acceleration corresponding to each candidate braking distance from the lane to the next lane; a calculation module configured to calculate a braking parameter allocation tree, including multiple branches connecting the current position and the target parking position, each branch containing multiple nodes, each node corresponding to a lane change position on one of the multiple lanes; an evaluation module configured to evaluate the lane change safety of each of the multiple branches; and an allocation module configured to allocate braking distances and lane change speeds to each node of the branch evaluated as having the highest lane change safety.
[0005] In one embodiment, determining the target parking location includes: calculating the shortest braking distance based on the current vehicle speed and the maximum braking deceleration; setting multiple candidate parking locations in the permitted parking area, using the location corresponding to the shortest braking distance as the starting point; scoring the parking safety of each candidate parking location; and determining the candidate parking location with the highest score as the target parking location.
[0006] In one embodiment, scoring the parking safety of each candidate parking location includes using one or more of the following scoring factors:
[0007] - The average deceleration calculated based on the distance between the candidate parking location and the vehicle's current location;
[0008] - The distance between the candidate parking location and the no-parking zones on the multiple lanes, optionally, the no-parking zones being areas near ramp entrances; and
[0009] - The safe parking priority of the sub-area where the candidate parking location is located, wherein the allowed parking area contains multiple sub-areas, each of which has a different pre-set safe parking priority.
[0010] In one embodiment, creating the solution space for the braking parameter combination includes: discretizing the lane change speed into m lane change speeds between its two boundary values, where m is an integer greater than 1; discretizing the lane change acceleration into n lane change accelerations between its two boundary values, where n is an integer greater than 1; obtaining m*n combinations based on the m lane change speeds and n lane change accelerations; and calculating the discretized braking distance corresponding to each combination based on the obtained m*n combinations, so as to create the solution space for the braking parameter combination.
[0011] In one embodiment, the two boundary values of the lane change speed are: 1) the current speed of the vehicle; 2) the speed determined based on the speeds of the vehicles in front and behind in the next lane; and the two boundary values of the lane change acceleration are predetermined by the vehicle's autonomous driving system.
[0012] In one embodiment, the control unit further includes a prediction module configured to: after calculating the braking distribution tree, use a supervised learning-based prediction model to predict: if a vehicle changes lanes at one of multiple nodes, the dynamic environmental information of the surrounding environment at that node when the vehicle changes lanes at that node, wherein the dynamic environmental information includes: the relative distance, relative speed, and relative acceleration between the vehicle and other surrounding vehicles when the vehicle changes lanes at that node.
[0013] In one embodiment, the prediction module is further configured to input static environmental information of the vehicle's surrounding environment into the prediction model, wherein the static environmental information includes: the number of lanes of the multi-lane road, the remaining drivable distance of each lane, and the distance between the vehicle and the no-parking area on the road. Optionally, the distance between the vehicle and the no-parking area on the road includes: the distance between the vehicle and the first ramp behind, the first ramp ahead, and the second ramp ahead, respectively.
[0014] In one embodiment, evaluating the lane change safety of each branch includes: using a reinforcement learning-based evaluation model, inputting the predicted dynamic environment information into the evaluation model to obtain the lane change safety score of each node on each branch; and calculating the sum of the scores of each node on each branch as the lane change safety evaluation result of that branch.
[0015] According to another aspect of the present invention, a vehicle is provided, comprising: a sensor for acquiring environmental information surrounding the vehicle; and a control unit as described above, communicatively connected to the sensor, configured to allocate braking distance and lane-changing speed to each lane of a plurality of lanes traversed during the parking process based on the environmental information, so as to control the vehicle to stop at a target parking position on a multi-lane road.
[0016] According to another aspect of the present invention, a control method for safe parking of a vehicle is provided, optionally executed by a control unit as described above and / or a vehicle as described above, the method comprising: determining a target parking position of the vehicle on a multi-lane road, wherein the distance between the vehicle's current position and the target parking position includes multiple braking distances along multiple lanes traversed by the vehicle during parking; creating a solution space for combining braking parameters for each of the multiple lanes, including multiple candidate braking distances for each lane, and a lane change speed and lane change acceleration corresponding to each candidate braking distance from the lane to the next lane; calculating a braking parameter allocation tree, including multiple branches connecting the current position and the target parking position, each branch containing multiple nodes, each node corresponding to a lane change position in one of the multiple lanes; evaluating the lane change safety of each branch among the multiple branches; and allocating braking distances and lane change speeds to each node of the branch evaluated as having the highest lane change safety.
[0017] According to another aspect of the invention, a machine-readable storage medium is provided that stores executable instructions, which, when executed, cause one or more processors to perform the control method described above.
[0018] The foregoing provides a summary of the main aspects of the invention in order to provide a basic understanding of these aspects and serves as an introduction to the detailed description that follows. Attached Figure Description
[0019] The technical solution of the present invention will become clearer from the following detailed description taken in conjunction with the accompanying drawings. It is to be understood that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.
[0020] Figure 1 The illustrations show some scenarios in which implementations of the present invention can be applied.
[0021] Figure 2 The illustration schematically depicts a vehicle equipped with a safe parking control function according to an embodiment of the present invention.
[0022] Figure 3 This is according to an embodiment of the present invention. Figure 2 A schematic block diagram of the vehicle's safe parking control unit.
[0023] Figure 4 This is a flowchart of a safe parking control process according to an embodiment of the present invention.
[0024] Figure 5 The schematic diagram illustrates the solution space of braking parameter combinations according to an embodiment of the present invention.
[0025] Figure 6 Schematic illustration based on Figure 5 Braking parameter allocation tree in the solution space of braking parameter combination.
[0026] Figure 7 A supervised learning-based prediction model according to an embodiment of the present invention is illustrated.
[0027] Figure 8 An evaluation model based on reinforcement learning according to an embodiment of the present invention is illustrated schematically.
[0028] Figure 9 This is a flowchart of a safe parking control method according to an embodiment of the present invention. Detailed Implementation
[0029] The specific embodiments of the present invention will now be described with reference to the accompanying drawings.
[0030] Embodiments of the present invention relate to a safe parking solution applicable to scenarios requiring multiple lane changes. For example, a vehicle needs to brake and change lanes two or more times on a multi-lane road before stopping at a safe location on that multi-lane road (i.e., the "target parking location" in the present invention).
[0031] Figure 1 Some scenarios in which implementations of the present invention can be applied are illustrated. For example... Figure 1 As shown, the current road has 5 lanes: lanes 1-4 and the emergency lane. Vehicle V (this vehicle) is traveling in lane 4. Other vehicles V1-V5 are traveling in lanes 2-4.
[0032] According to an embodiment of the present invention, multiple permitted parking areas on a lane are pre-prioritized, see [link / reference]. Figure 1The diagram shows multiple permitted parking areas located on the emergency lane and lane 1, each with a pre-defined safe parking priority. Figure 1 The priority levels are categorized as "Priority 1-Priority 4". Higher parking safety levels correspond to higher parking priority. For example, if both Priority 1 and Priority 2 areas are potential parking areas, the vehicle should be parked in the Priority 1 area first. Additionally, there may be no-parking zones on the road; see [link to relevant documentation]. Figure 1 The "No Parking" zones are designated areas. These zones may be near ramp entrances or areas where lane changes are prohibited.
[0033] According to embodiments of the present invention, a target parking position on a multi-lane road can be determined by considering both feasibility and safety factors. According to embodiments of the present invention, an optimal lane-changing branch (i.e., the branch with the lowest cost or the branch with the highest lane-changing safety) can also be determined from multiple branches of the braking parameter allocation tree connecting the vehicle's current position and the target parking position. According to embodiments of the present invention, the braking distance and lane-changing speed for each lane along the multiple lanes traversed by the vehicle can also be rationally allocated.
[0034] Embodiments of the present invention can be implemented using machine learning models, such as supervised learning-based prediction models, reinforcement learning-based evaluation models, and hybrid learning models.
[0035] Understandable Figure 1 This is merely an illustrative scenario, and the application scenarios of this invention are not limited to these examples. Figure 1 The scenario can be as follows. For example, the number of lanes can be more than two. The number of traffic participants in the traffic environment can be more or less, and other types of traffic parameters can be included, such as animals or cyclists. Embodiments of the present invention are applicable to roads that are multi-lane highways or multi-lane urban roads.
[0036] Figure 2 A vehicle V according to an embodiment of the present invention is shown, which includes a sensor 10 and a safe parking control unit 20 (hereinafter referred to as "control unit 20") communicatively connected to the sensor 10. The sensor 10 and the control unit 20 together constitute a safe parking system 100. The vehicle V equipped with the safe parking system 100 has a safe parking control function.
[0037] Sensor 10 may include vehicle state sensors and environmental sensors. Vehicle state sensors are used to sense the state of vehicle V, such as its speed and acceleration. Vehicle state sensors may include wheel speed sensors, acceleration sensors, etc. Environmental sensors are used to sense the surrounding environment of vehicle V, such as the current road conditions and the state information of other vehicles around vehicle V. Environmental sensors may include radar (e.g., lidar, ultrasonic radar, millimeter-wave radar) and cameras (e.g., single-target, multi-target, surround-view). Cameras can obtain environmental information around vehicle V through image or video analysis, such as the relative distance between the vehicle and the roadside or obstacles. Radars can obtain the relative distance between the vehicle and the roadside or obstacles through point cloud analysis. Environmental sensors may include multiple environmental sensors disposed around the vehicle body, and this arrangement considers safety redundancy, i.e., ensuring that the environmental conditions around the vehicle can be fully collected.
[0038] In addition, when vehicle V is in a vehicle-to-everything (V2X) scenario, sensor 10 may also include an Internet of Things (IoT) sensor for receiving information from roadside units, cloud servers, and other vehicles.
[0039] The control unit 20 determines static and dynamic environmental information based on the sensing results of the sensor 10.
[0040] Static environment information includes: the number of lanes in the road currently being traveled by vehicle V, and the remaining drivable distance in each lane. In scenarios where there are ramps on the current road, static environment information may also include the longitudinal distance between vehicle V and the ramp, such as the distance between vehicle V and the first and second ramps in front of it, and the distance between vehicle V and the first ramp behind it.
[0041] Dynamic environment information includes: the relative distance, relative velocity, and relative acceleration between the vehicle V and surrounding objects (e.g., moving and stationary objects). The relative distance may include lateral and longitudinal relative distances. The relative velocity may include lateral and longitudinal relative velocities. The relative acceleration may include lateral and longitudinal relative accelerations.
[0042] Figure 3 One implementation of the control unit 20 is shown. For example... Figure 3As shown, the control unit 20 includes: a determination module 21, a creation module 22, a calculation module 23, a prediction module 24, an evaluation module 25, and an allocation module 26. It is understood that the naming of these modules is based on function and not intended to limit their physical location or implementation. For example, these modules may be located on the same chip or circuit, or on different chips or circuits. These modules may also be combined into fewer modules based on function, or divided into more modules based on function.
[0043] The control unit 20 and its modules 21-26 can be implemented in hardware, software, or a combination of both. For the hardware implementation, it can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), data signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, electronic units designed to perform their functions, or combinations thereof. For the software implementation, it can be implemented using microcode, program code, or code segments, and can also be stored in a machine-readable storage medium such as a storage component.
[0044] In one embodiment, the control unit 20 may be implemented including a memory and a processor. The memory stores a traffic regulations database, multiple information categories, and instructions that, when executed by the processor, cause the processor to perform a safe stop control method according to an embodiment of the present invention.
[0045] In one embodiment, the control unit 20 is located in the ECU of the vehicle V or in the domain controller of the vehicle V.
[0046] Figure 4 A control process 400 for safe parking according to an embodiment of the present invention is shown. This control process 400 can be executed by means of the aforementioned control unit 20, and therefore the related description above also applies. For clarity, the control unit 20 controlling the vehicle V in... Figure 1 The following explanation will be based on the implementation of control process 400 in a specific scenario.
[0047] See Figure 4 In box 410, the determining module 21 determines the target parking position of vehicle V on a multi-lane road. The distance between the current position of vehicle V and the determined target parking position is the total braking distance. This total braking distance includes multiple braking distances along the multiple lanes traversed by vehicle V during the parking process. For example, see... Figure 1 If the current position of vehicle V is P0 and the target parking position is determined to be P4, then the total braking distance includes the braking distances in lane 4, lane 3, lane 2, and lane 1 respectively.
[0048] In one embodiment, determining the target parking location in block 410 can be achieved by selecting the safest parking location from multiple candidate parking locations. The specific implementation of this embodiment is described below.
[0049] In box 411, the shortest braking distance is calculated based on the current speed and maximum braking deceleration of vehicle V. That is, the shortest travel distance for vehicle V to brake and decelerate from its current speed to zero and stop using the maximum braking deceleration.
[0050] In box 412, multiple candidate parking positions are set in the permitted parking area, with the position corresponding to the shortest braking distance as the starting point.
[0051] See Figure 1 To illustrate, box 412 shows the setting of multiple candidate parking locations. For example... Figure 1 As shown, the current position of vehicle V is P0, and the parking position corresponding to the shortest braking distance is P1. Starting from this position P1, multiple positions P2-P4 are set backward at equal intervals (for example, the interval can be predetermined) within the allowed parking area. These multiple positions, together with the starting position, become multiple candidate parking positions P1-P4.
[0052] In box 413, a parking safety score is given for each candidate parking location.
[0053] In one embodiment, the following three scoring factors F1-F3 can be used to score parking safety.
[0054] F1: The average deceleration calculated based on the distance between the candidate parking position and the vehicle's current position. The smaller the average deceleration, the higher the first parking safety score S1 obtained based on this scoring factor F1. For example, four average decelerations are calculated based on the distances between the vehicle V's current position P0 and candidate parking positions P1-P4. The average deceleration calculated based on P0 and P4 is the smallest, so position P4 obtains the highest first safety score S1 based on this scoring factor F1.
[0055] F2: The distance between the candidate parking location and a no-parking zone on the road. This no-parking zone could be, for example, a ramp entrance. The greater the distance between the candidate parking location and the no-parking zone, the higher the second parking safety score S2 derived from this scoring factor F2. For example, among multiple candidate parking locations P1-P4, location P4 is the farthest from the ramp entrance, therefore location P4 will receive the highest second parking safety score based on this scoring factor F2.
[0056] F3: Safe parking priority of the sub-region where the candidate parking location is located. The higher the safe parking priority (i.e., parking priority) of the sub-region where the candidate parking location is located, the higher the third parking safety score S3 obtained based on this scoring factor F3. For example, among multiple candidate parking locations P1-P4, location P4 has the highest safe parking priority, so location P4 has the highest third parking safety score based on this scoring factor F3.
[0057] In this embodiment, the parking safety score for each candidate parking location can be the sum of the first to third parking safety scores described above. Alternatively, the parking safety score for each candidate parking location can be calculated using the following formula:
[0058] S = k1*S1 + k2*S2 + k3*S3,
[0059] Where S is the safety score of the candidate parking location;
[0060] S1, S2, and S3 are the first to third parking safety scores mentioned above, respectively.
[0061] k1-k3 are the weights of the first to third parking safety scores mentioned above, and satisfy: 0≤k1≤1, 0≤k2≤1, 0≤k3≤1, k1+k2+k3=1.
[0062] k1-k3 can be set based on specific application scenarios. For example, on roads with many ramps, k2 can be set relatively high; when users have high requirements for comfort, k1 can be set relatively high.
[0063] In box 414, the candidate parking location with the highest parking safety score is determined as the target parking location.
[0064] See also Figure 4 In box 420, module 22 creates a solution space for the braking parameters of the current lane, which includes multiple candidate braking distances on the current lane, as well as the lane change speed and lane change acceleration corresponding to each candidate braking distance when changing lanes from the current lane to the next lane.
[0065] In one embodiment, the creation of the braking parameter combination solution space in block 420 can be achieved by discretizing multiple lane change speeds and multiple lane change accelerations and combining these discretized values. The specific implementation of this embodiment is described below.
[0066] In box 421, the lane change speed is discretized into m lane change speeds between its two boundary values, where m is an integer greater than 1. The two boundary values for the lane change speed are: 1) the current speed of vehicle V; and 2) the speed determined based on the speeds of the vehicles in front and behind in the next lane.
[0067] In box 422, the lane change acceleration is discretized into n lane change accelerations between its two boundary values, where n is an integer greater than 1. The two boundary values of the lane change acceleration are predetermined by the autonomous driving system of vehicle V.
[0068] In box 423, based on m lane change speeds and n lane change accelerations, m*n combinations containing lane change speeds and lane change accelerations are obtained.
[0069] In box 424, the candidate braking distances on the current lane corresponding to each combination are calculated to create the solution space of the braking parameter combination.
[0070] In box 430, the calculation module 23 calculates the braking parameter allocation tree, which contains multiple branches connecting the current position of vehicle V with the target parking position. Each branch contains multiple nodes, and each node corresponds to the lane change position in each of the multiple lanes that vehicle V passes through.
[0071] Figure 5 The solution space of braking parameter combinations (hereinafter referred to as "combination solution space") is illustrated schematically. For example, the combination solution space α for lane 4, the combination solution space β for lane 3, etc. Figure 6 Schematic illustration based on Figure 5 The braking parameter allocation tree (hereinafter referred to as "allocation tree") is used in the solution space of braking parameter combinations. See below. Figure 5 and Figure 6 An embodiment of boxes 420 and 430 is described.
[0072] The lane change acceleration boundary is the boundary set by the autonomous driving system of vehicle V.
[0073] For non-emergency lanes, the maximum speed for lane changing is the current speed v, that is: The lower limit for lane-changing speed is the speed of a moving obstacle in the next lane when the vehicle changes lanes to the next lane:
[0074]
[0075] Among them, v o It is the current or predicted speed of the moving obstacle located in the next lane, a o It is the current or predicted acceleration of the moving obstacle in the next lane, and t is the default lane change time.
[0076] For example, if the current lane is lane 4, the solution space α for the combination of braking distance and lane change speed corresponding to the current lane can be determined by the following steps:
[0077] Lane change speed limit is Lane change acceleration boundary is Within the boundary value range, the lane change speed and lane change acceleration are discretized separately to obtain the discretized sequence. and Next, the discretized sequence of braking distances is obtained according to the following formula. Elements in:
[0078]
[0079] Where v is the initial speed when entering the lane change lane (i.e., the next lane after the current lane), v e yes For a certain element, a e yes One of the elements, therefore The number of elements is sm×sn. The speed v of a vehicle changing lanes... e The corresponding braking distance sequence of size sn All lane change speeds and their corresponding braking distance sequences together constitute the solution space α of the combination of braking distance and lane change speed.
[0080] In this embodiment of the invention, discretization sampling refers to selecting a certain lane-changing speed v from the solution space α of the combination of braking distance and lane-changing speed. e and its corresponding braking distance sequence a certain element The sampling point a formed, i.e.
[0081] Next, starting from a sampling point a in the combined solution space α, the braking parameter allocation tree T with a as the root node is obtained. a .
[0082] Braking distance and lane change speed distribution tree T a It is a tree with root node 'a' and height equal to the total number of lanes traversed by vehicle 'V'.
[0083] Allocation tree T a The following steps can be used to obtain it:
[0084] Using a sampling point 'a' in the combined solution space α as the initial condition for vehicle V to change lanes to the next lane, we obtain the combined solution space β for the next lane. This process is repeated until the last lane (e.g., the emergency lane), thus obtaining the allocation tree T of braking distance and lane change speed. a .
[0085] See Figure 5 Vehicle V is currently in lane 4, and its corresponding combinatorial solution space α is composed of nodes. Composition. Let the node Let sampling point a be the initial condition for changing lanes to the next lane. This yields the combined solution space β of vehicle V in lane 3. Then, sampling points are taken from the combined solution space β. As the initial condition for the controlled vehicle to change lanes to lane 2, the corresponding solution space γ is obtained. Similarly, the solution space δ for lane 1 and the solution space ε for the emergency lane are obtained. See also... Figure 6 Taking sampling point a as the root node, we can obtain a tree of height 5, which is the assignment tree T corresponding to sampling point a, containing braking distance and lane change speed. a .
[0086] See also Figure 4 In box 440, evaluation module 24 evaluates the lane change safety of each of the multiple branches.
[0087] In one embodiment, a prediction model based on supervised learning and an evaluation model based on reinforcement learning are used to obtain the evaluation results of lane change safety. The specific implementation of this embodiment is described below.
[0088] In box 441, prediction module 24 uses a supervised learning-based prediction model to predict the dynamic environment information of the surrounding environment at a node when vehicle V changes lanes in the current lane. This dynamic environment information may include the relative distance, relative speed, and relative acceleration between the vehicle and other vehicles at the node when it changes lanes.
[0089] In one embodiment, static environmental information about the vehicle's surroundings is input into a prediction model, and the prediction model outputs the predicted information, i.e., the predicted dynamic environmental information. The input static environmental information may include: the number of lanes, the remaining drivable distance for each lane (if the remaining drivable distance for a lane is very long, for example, exceeding a threshold distance, a predetermined default value, such as 3 km, can be used to represent the remaining drivable distance for that lane), and the distance between the vehicle and no-parking zones on the road. For example, the distance between the vehicle's current position and the first ramp behind it, the first ramp ahead it, and the second ramp ahead it.
[0090] Figure 7 An example of a prediction model based on supervised learning is shown. For example... Figure 7As shown, the prediction model is structured as a neural network consisting of N fully connected layers. The input includes static and dynamic environmental information of vehicle V when it is in the current lane, such as: vehicle speed, acceleration, relative speed and relative relative speed between vehicle V and obstacles in the current lane, and the number of lanes. The output is the predicted dynamic environmental information of vehicle V after it changes lanes, such as the lateral / longitudinal distance, lateral / longitudinal relative speed, and lateral / longitudinal acceleration between vehicle V and surrounding moving obstacles when it changes lanes to the next lane.
[0091] For example, the following steps can be used to "predict the dynamic environment information of each node in tree Ta". From T a A path to leaf node k in the middle Starting from layer 1, iteratively call the prediction model and traverse downwards. Each node in the middle is obtained The dynamic environment information corresponding to each node.
[0092] In box 442, a reinforcement learning-based evaluation model is used to score the lane change safety of each node on each branch road by using the predicted dynamic environment information.
[0093] In box 443, the sum of the scores of each node on each branch is calculated as the lane change safety assessment result for that branch.
[0094] Figure 8 An example of an evaluation model based on reinforcement learning is shown. Figure 8 As shown, the evaluation model is structured as a neural network consisting of N fully connected layers. The inputs include static and dynamic environmental information of vehicle V in the current lane, and the output is the evaluated cost value. A lower cost value indicates a more optimal input lane change speed and braking distance. In one embodiment, the output can also be a safety score; a higher safety score indicates a more optimal input lane change speed and braking distance.
[0095] In one embodiment, the allocation tree T can be evaluated using the following steps. a The cost of each node. For example... Figure 6 As shown, from Starting from the 5th layer, iteratively call the evaluation model and traverse upwards. Each node in the middle is obtained The cost corresponding to each node is used to obtain the final path. The corresponding cost.
[0096] See also Figure 4 In box 450, the allocation module 26 allocates braking distance and lane change speed to each node of the branch that is assessed as having the highest lane change safety.
[0097] In one embodiment, the following steps can be used to select the branch with the highest lane-changing safety (i.e., the branch with the lowest cost) and allocate braking distance and lane-changing speed. Selecting the branch with the lowest cost... As the optimal path, where a′ is a sampling point in the combined solution space α, and k′ is the braking distance and lane change speed allocation tree T corresponding to the root node a′. a′ For a leaf node k′, according to the path The distances to each node and the lane-changing speeds determine the braking distances and corresponding lane-changing speeds along the route from the current lane to the last lane.
[0098] Therefore, embodiments of the present invention provide a safe parking solution based on supervised learning and reinforcement learning models. During multiple lane changes by the vehicle, a combined solution space of braking distance and lane change speed is determined based on the lane change speed and lane change acceleration boundaries, and discretized and sampled as the initial state for the vehicle to change lanes to the next lane, thus obtaining the combined solution space for the next lane, until the last lane. Starting from a sampling point in the combined solution space of the current lane, the sampling points in the combined solution spaces of each lane are traversed downwards, and the prediction model and evaluation model are called multiple times to determine the lane change safety score of each sampling point in the combined solution space of the current lane, thereby determining the braking distance and corresponding lane change speed in each lane the vehicle passes through during the parking process.
[0099] According to embodiments of the present invention, reinforcement learning and supervised learning are introduced to address the problems of traditional algorithms being mechanically rigid, unable to adapt to dynamic environments, and having low robustness. Furthermore, an innovative solution based on model-based iterative scoring and discretization sampling is proposed to reduce the training difficulty of machine learning models and improve their generalization ability.
[0100] Figure 9 A control method 900 for safe parking of a vehicle according to an embodiment of the present invention is shown. This method 900 can be executed by the aforementioned control unit 20, and therefore the above description of the control unit 20 also applies here.
[0101] See Figure 9 In step S910, the target parking position of the vehicle on the target parking lane of the multi-lane road is determined, wherein the distance between the current position of the vehicle and the target parking position is the total braking distance, which includes multiple braking distances of the multiple lanes that the vehicle passes through during the parking process.
[0102] In step S920, a solution space for the braking parameters of the current lane is created, which includes multiple candidate braking distances in the current lane, as well as the lane change speed and lane change acceleration corresponding to each candidate braking distance when changing lanes from the current lane to the next lane.
[0103] In step S930, a braking parameter allocation tree is calculated, which contains multiple branches connecting the vehicle's current position and the target parking position. Each branch contains multiple nodes, and each node corresponds to a lane change position in one of the multiple lanes traversed.
[0104] In step S940, the lane change safety of each of the multiple branches is evaluated.
[0105] In step S950, braking distance and lane change speed are assigned to each node of the branch with the highest lane change safety for evaluation.
[0106] The present invention also provides a machine-readable storage medium storing executable instructions that, when executed, cause the machine to perform the control method 900 as described above.
[0107] It is understood that all operations in the processes and methods described above are merely exemplary, and the present invention is not limited to any operation in the methods or the order of such operations, but should be covered by all other equivalent transformations under the same or similar concept.
[0108] It is understood that a control unit may include one or more processors. These processors may be implemented using electronic hardware, computer software, or any combination thereof. Whether these processors are implemented as hardware or software will depend on the specific application and the overall design constraints imposed on the system. As an example, the processors, any portions of processors, or any combination of processors provided in this invention may be implemented as microprocessors, microcontrollers, digital signal processors (DSPs), field-programmable gate arrays (FPGAs), programmable logic devices (PLDs), state machines, gate logic, discrete hardware circuits, and other suitable processing units configured to perform the various functions described in this invention. The functionality of the processors, any portions of processors, or any combination of processors provided in this invention may be implemented as software executed by a microprocessor, microcontroller, DSP, or other suitable platform.
[0109] Software can be broadly considered as representing instructions, instruction sets, code, code segments, program code, programs, subroutines, software modules, applications, software applications, software packages, routines, subroutines, objects, running threads, procedures, functions, etc. Software can reside on a computer-readable medium. Computer-readable media can include, for example, memory, which can be, for example, magnetic storage devices (e.g., hard disks, floppy disks, magnetic stripes), optical disks, smart cards, flash memory devices, random access memory (RAM), read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), registers, or removable disks. Although memory is shown as separate from the processor in several aspects of the invention, memory can also reside within the processor (e.g., in caches or registers).
[0110] The above description is provided to enable any person skilled in the art to implement the various aspects described herein. Various modifications to these aspects will be apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects. Therefore, the claims are not intended to be limited to the aspects shown herein. All structural and functional equivalents of the elements of the various aspects described herein that are known or would be apparent to those skilled in the art are expressly incorporated herein by reference and are intended to be covered by the claims.
Claims
1. A control unit for safely parking a vehicle, comprising: The determination module is configured to determine the target parking position of the vehicle on a multi-lane road, wherein the distance between the vehicle's current position and the target parking position includes multiple braking distances of the vehicle along multiple lanes during the parking process; The module is configured to create a solution space for the combination of braking parameters for each of the multiple lanes, which includes multiple candidate braking distances on each lane, and the lane change speed and lane change acceleration corresponding to each candidate braking distance when changing lanes from that lane to the next lane. The calculation module is configured to calculate the braking parameter allocation tree, which includes multiple branches connecting the current position and the target parking position. Each branch contains multiple nodes, and each node corresponds to the lane change position in one of the multiple lanes. An evaluation module is configured to evaluate the lane change safety of each of the multiple branch roads; and an allocation module is configured to allocate braking distance and lane change speed to each node of the branch road that is evaluated as having the highest lane change safety. Determining the target parking location includes: The shortest braking distance is calculated based on the vehicle's current speed and maximum braking deceleration; multiple candidate parking positions are set in the permitted parking area, using the position corresponding to the shortest braking distance as the starting point. Each candidate parking location is scored for parking safety; and the candidate parking location with the highest score is determined as the target parking location.
2. The control unit as claimed in claim 1, wherein, Scoring the parking safety of each candidate parking location involves using one or more of the following scoring factors: The average deceleration is calculated based on the distance between the candidate parking location and the vehicle's current location; The distance between the candidate parking location and the no-parking zones on the multiple lanes, wherein the no-parking zones are areas near the ramp entrances; and the safe parking priority of the sub-area where the candidate parking location is located, wherein the permitted parking area includes multiple sub-areas, each with a pre-set different safe parking priority.
3. The control unit as described in any one of claims 1-2, wherein, Creating the solution space for the braking parameter combination includes: The lane change speed is discretized into m lane change speeds between its two boundary values, where m is an integer greater than 1; The lane change acceleration is discretized into n lane change accelerations between its two boundary values, where n is an integer greater than 1; Based on m lane change speeds and n lane change accelerations, m*n combinations are obtained; and based on the obtained m*n combinations, the discretized braking distance corresponding to each combination is calculated to create the solution space of the braking parameter combinations.
4. The control unit as claimed in claim 3, wherein, The two boundary values for the lane change speed are: 1) the current speed of the vehicle; 2) the speed determined based on the speeds of the vehicles in front and behind in the next lane; and wherein the two boundary values for the lane change acceleration are predetermined by the vehicle's autonomous driving system.
5. The control unit as claimed in claim 1, wherein, The control unit further includes a prediction module, which is configured to: after calculating the braking parameter allocation tree, use a supervised learning-based prediction model to predict: if the vehicle changes lanes at one of the multiple nodes, the dynamic environmental information of the surrounding environment of the node when the vehicle changes lanes at that node, wherein the dynamic environmental information includes: the relative distance, relative speed and relative acceleration between the vehicle and other surrounding vehicles when the vehicle changes lanes at that node.
6. The control unit as claimed in claim 5, wherein, The prediction module is further configured to input static environmental information of the vehicle's surrounding environment into the prediction model, so as to output predicted dynamic environmental information from the prediction model. The static environmental information includes: the number of lanes of the multi-lane road, the remaining drivable distance of each lane, and the distance between the vehicle and the no-parking area on the road. The distance between the vehicle and the no-parking area on the road includes: the distance between the vehicle and the first ramp behind, the first ramp ahead, and the second ramp ahead, respectively.
7. The control unit as claimed in claim 1, wherein, Assessing the lane change safety of each branch road includes: An evaluation model based on reinforcement learning is adopted. The predicted dynamic environment information is input into the evaluation model to obtain the lane change safety score of each node on each branch; and the sum of the scores of each node on each branch is calculated as the lane change safety evaluation result of that branch.
8. A vehicle comprising: Sensors are used to collect information about the environment around the vehicle; The control unit, as described in any one of claims 1-7, is communicatively connected to the sensor and configured to allocate braking distance and lane-changing speed to each lane of the multiple lanes the vehicle passes through during the parking process based on the environmental information, so as to control the vehicle to stop at the target parking position on the multi-lane road.
9. A control method for safely parking a vehicle, executed by a control unit as described in any one of claims 1-7 and / or a vehicle as described in claim 8, the method comprising: Determine the target parking position of the vehicle on a multi-lane road, wherein the distance between the vehicle's current position and the target parking position includes multiple braking distances of the vehicle along multiple lanes during the parking process; Create a solution space for the braking parameters of each lane in the multiple lanes, which includes multiple candidate braking distances in each lane, and the lane change speed and lane change acceleration corresponding to each candidate braking distance when changing lanes from that lane to the next lane. Calculate the braking parameter allocation tree, which includes multiple branches connecting the current position and the target parking position. Each branch contains multiple nodes, and each node corresponds to a lane change position in one of the multiple lanes. Assess the lane change safety of each of the multiple branch roads; and assign braking distance and lane change speed to each node of the branch road that is assessed as having the highest lane change safety.
10. A machine-readable storage medium storing executable instructions that, when executed, cause one or more processors to perform the control method of claim 9.
Citation Information
Patent Citations
Method for guiding vehicle to run, related system and storage medium
CN113619602A
Driving control apparatus for vehicle
EP3974277A1
