Path planning method and system based on indoor intelligent vehicle

Abstract

The invention provides a path planning method and system based on an indoor intelligent vehicle, and the method comprises the steps: obtaining indoor map recognition information; indoor map identification information is converted into target map identification information of a Frenet space coordinate system, and a potential field is introduced based on an improved artificial potential field algorithm; carrying out real-time scanning through a preset laser radar on the indoor intelligent vehicle, and generating a candidate track; and updating the memory list in real time according to a track evaluation result, and controlling the indoor intelligent vehicle to run according to an updating result and continuously scan. According to the path planning method and system based on the indoor intelligent vehicle, indoor map identification information is converted into a Frenet space coordinate system, and an artificial potential field algorithm is improved; the laser radar is controlled to scan in real time, the optimal track with the optimal evaluation is screened to update the memory list in real time, the indoor intelligent vehicle runs based on the optimal track after updating, track selection is more efficient, and running stability is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of path planning, and in particular to a path planning method and system based on an indoor intelligent vehicle. Background Art

[0002] Intelligent networked vehicles, also known as smart cars, are high-tech vehicles that integrate high-end sensors, control units, actuators and other equipment. They use modern communication and network technologies to achieve the interaction and sharing of intelligent information, thus having the ability to make autonomous decisions and control. This type of vehicle can effectively exchange data with the external environment through advanced technical means to ensure that it can make intelligent judgments and responses during driving. At present, smart cars are widely used in factories. For example, car factories use a large number of smart cars to transport auto parts. The process of smart car movement involves path planning. At present, the trajectory planning of smart cars is mostly global static planning. When facing the complex and changeable factory environment, there are problems such as low security and poor stability.

[0003] In view of this, there is an urgent need for a path planning method and system based on an indoor smart car to at least solve the above-mentioned shortcomings. Summary of the invention

[0004] One of the purposes of the present invention is to provide a path planning method and system based on an indoor intelligent car, which converts the recognized indoor map recognition information into the Frenet space coordinate system to obtain the target map recognition information; based on the improved artificial potential field algorithm and according to the target map recognition information, a reference line potential field, a velocity potential field and an obstacle repulsion potential field are introduced; the laser radar is controlled to scan in real time, and the optimal trajectory with the best evaluation is screened to update the memory list in real time. After the update, the indoor intelligent car drives based on the optimal trajectory and then continues to scan for local trajectory planning. The trajectory selection is efficient and the driving stability is higher.

[0005] An embodiment of the present invention provides a path planning method based on an indoor smart car, comprising:

[0006] Step 1: Obtain indoor map identification information;

[0007] Step 2: Convert the indoor map recognition information into the target map recognition information of the Frenet space coordinate system, and introduce the reference line potential field, velocity potential field and obstacle repulsion potential field;

[0008] Step 3: Use the laser radar preset on the indoor smart car to perform real-time scanning to generate candidate trajectories;

[0009] Step 4: The memory list is updated in real time according to the trajectory evaluation results of the candidate trajectories, and the smart car in the control room drives and continues scanning according to the updated results.

[0010] Preferably, step 1: obtaining indoor map identification information includes:

[0011] Based on the preset lidar and IMU sensors on the indoor smart car, indoor map recognition information is obtained.

[0012] Preferably, step 2: converting the indoor map recognition information into target map recognition information of the Frenet space coordinate system, and introducing a reference line potential field, a velocity potential field and an obstacle repulsion potential field, comprises:

[0013] Get vehicle parameters;

[0014] According to the vehicle parameters, the indoor map recognition information in the Cartesian coordinate system is converted into the target map recognition information in the Frenet space coordinate system;

[0015] According to the vehicle parameters, the artificial potential field algorithm is improved, and the improved artificial potential field algorithm includes: adding reference line potential field, velocity potential field and obstacle repulsion potential field.

[0016] Preferably, step 3: performing real-time scanning by a laser radar preset on the indoor smart car to generate candidate trajectories includes:

[0017] Real-time scanning is performed through the preset laser radar on the indoor smart car, and multiple candidate trajectories are planned based on dynamic obstacle avoidance trajectory planning technology.

[0018] Preferably, a laser radar preset on the indoor smart car is used for real-time scanning, and multiple candidate trajectories are planned based on the dynamic obstacle avoidance trajectory planning technology, including:

[0019] According to the real-time scanning results, try to obtain the suspicious point cloud set;

[0020] If the acquisition attempt is successful, the target identification set is determined;

[0021] Obtain scanning parameters of the suspicious point cloud set;

[0022] According to the scanning parameters, a scanning point cloud set of the target identification object in the target identification object set is obtained;

[0023] Perform point cloud matching on the suspicious point cloud set and the scanned point cloud set to determine the matching identification object;

[0024] Obtain an actual point cloud set of the matching identification object corresponding to the scanning parameters;

[0025] Perform dynamic obstacle avoidance and plan candidate trajectories based on the actual point cloud set;

[0026] Wherein, obtaining a scanning point cloud set of an identification object in an identification object set according to the scanning parameters includes:

[0027] Obtain historical scanning parameters of the historical scanning point cloud set of the target object, and if the historical scanning parameters match the scanning parameters, use the corresponding historical scanning point cloud set as the scanning point cloud set;

[0028] and / or,

[0029] A scanning simulation of the target object is performed according to the scanning parameters, and a simulated point cloud set obtained by the scanning simulation is used as a scanning point cloud set.

[0030] Preferably, determining the target identifier set includes:

[0031] Try to obtain scan confirmation objects within the preset target range of the suspicious point cloud set;

[0032] If the acquisition attempt is successful, a relationship feature set is extracted based on the scanned confirmed object and the suspicious point cloud set, and the relationship features include: the type of the scanned confirmed object and the distance between the scanned confirmed object and the suspicious point cloud set;

[0033] According to the spatial distribution of the identification object set in the space where the indoor smart car is located, a relational feature set library is constructed;

[0034] Determine a target identification object set according to the relationship feature set and the relationship feature set library;

[0035] If the acquisition attempt fails, the identification object set of the space where the indoor smart car is located is used as the target identification object set.

[0036] Preferably, step 4: updating the memory list in real time according to the trajectory evaluation results of the candidate trajectories, and controlling the indoor smart car to drive and continue scanning according to the updated results, includes:

[0037] Candidate trajectories are evaluated based on weighted assessment of safety loss and comfort, acceleration check, curvature check, speed check, deviation from reference line index, and dynamic obstacle distance evaluation index;

[0038] According to the trajectory evaluation results of the candidate trajectories, an optimal trajectory, a collision trajectory and a suboptimal trajectory are determined;

[0039] Put the optimal trajectory into the memory list, and put the collision trajectory and suboptimal trajectory into the forget list;

[0040] The indoor smart car is controlled to drive and continue scanning according to the optimal trajectory updated in real time according to the memory list.

[0041] An embodiment of the present invention provides a path planning method based on an indoor smart car, further comprising:

[0042] When the smart car in the control room drives and continues scanning according to the update results, the relay scanning data is corrected to obtain the target relay scanning data.

[0043] Preferably, when the smart car in the control room drives and continues scanning according to the update result, the relay scanning data is corrected to obtain the target relay scanning data, including:

[0044] Get the scanning range information of the indoor smart car;

[0045] Determine the trajectory points in the updated optimal trajectory, and determine the target tangent lines of the trajectory points;

[0046] Visualize the scanning range information and determine the process coverage area based on the scanning range information and target tangent;

[0047] Obtain the target coverage area of ​​the process coverage area during the driving process of the indoor smart car based on the updated optimal trajectory;

[0048] Determine the end point of the updated optimal trajectory;

[0049] Determine the relay scanning area based on the endpoint trajectory point and scanning range information;

[0050] Determine the overlapping area between the target coverage area and the relay scanning area;

[0051] Obtain the corresponding information of the scanning points in the overlapped area and the relay scanning area;

[0052] According to the corresponding information of the scanning points, the fusion relationship of the scanning data corresponding to the scanning points is determined;

[0053] The scanning data is fused according to the fusion relationship to obtain the target relay scanning data.

[0054] The embodiment of the present invention provides a path planning method based on an indoor smart car, further comprising:

[0055] Obtaining task evaluation information after the indoor smart car driving task is completed;

[0056] Parse the task evaluation information and obtain the first negative evaluation semantics of the indoor smart car;

[0057] Obtaining the second negative evaluation semantics of the negative evaluation type preset by the indoor smart car;

[0058] Performing semantic matching on the first negative evaluation semantics and the second negative evaluation semantics to determine a target negative evaluation type that matches the semantic match and its corresponding third negative evaluation semantics;

[0059] When the target negative evaluation type is inaccurate obstacle recognition, the obstacle type is determined according to the third negative evaluation semantics;

[0060] Setting scanning scenes of target obstacles of different obstacle types and multiple placement states for the preset target detection model of the indoor smart car to learn;

[0061] When the target negative evaluation type is poor passability, the target pass height set is determined according to the third negative evaluation semantics;

[0062] Get the theoretical passing height of indoor smart cars;

[0063] Calculate the number of occurrences of target passing heights that are lower than the theoretical passing height;

[0064] If the number of occurrences is greater than or equal to the number of occurrences threshold, the third negative evaluation semantics corresponding to the target passing height lower than the theoretical passing height is used as the fourth negative evaluation semantics;

[0065] Based on the preset passability impact basis extraction template and the fourth negative evaluation semantics, determine the passability impact basis;

[0066] According to the influence of passability, the passability of indoor smart car is optimized.

[0067] An embodiment of the present invention provides a path planning system based on an indoor smart car, comprising:

[0068] Indoor map identification information acquisition subsystem, used to acquire indoor map identification information;

[0069] The coordinate conversion subsystem is used to convert the indoor map recognition information into the target map recognition information of the Frenet space coordinate system, and introduce the reference line potential field, velocity potential field and obstacle repulsion potential field;

[0070] The candidate trajectory generation subsystem is used to generate candidate trajectories by real-time scanning using a laser radar preset on the indoor smart car;

[0071] The continuous scanning subsystem is used to update the memory list in real time according to the trajectory evaluation results of the candidate trajectories, and control the indoor smart car to drive and continue scanning according to the updated results.

[0072] The beneficial effects of the present invention are:

[0073] The present invention converts the recognized indoor map recognition information into the Frenet space coordinate system to obtain the target map recognition information; based on the improved artificial potential field algorithm and according to the target map recognition information, the reference line potential field, the velocity potential field and the obstacle repulsion potential field are introduced; the laser radar is controlled to scan in real time, the optimal trajectory with the best evaluation is screened and the memory list is updated in real time, and after the update, the indoor smart car drives based on the optimal trajectory and then continues to scan to perform local trajectory planning, so that the trajectory selection is highly efficient and the driving stability is higher.

[0074] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the present application documents.

[0075] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0077] Figure 1 A schematic diagram of a path planning method based on an indoor smart car in an embodiment of the present invention;

[0078] Figure 2 Schematic diagram of a path planning system based on an indoor smart car in an embodiment of the present invention. DETAILED DESCRIPTION

[0079] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0080] The embodiment of the present invention provides a path planning method based on an indoor smart car, such as Figure 1 As shown, including:

[0081] Step 1: Obtain indoor map recognition information; the indoor map recognition information is: spatial information such as indoor obstacles and indoor layout initially recognized based on the laser radar and IMU;

[0082] Step 2: Convert the indoor map recognition information into the target map recognition information of the Frenet space coordinate system, and introduce the reference line potential field, velocity potential field and obstacle repulsion potential field; wherein, the indoor map recognition information is the indoor map recognition information in the Cartesian coordinate system; the target map recognition information is the indoor map recognition information in the Frenet space coordinate system; the reference line potential field is a virtual force field used to guide the vehicle to travel along a predetermined path or reference line, which is manifested as an attraction to the vehicle to keep it on the path, and the reference line potential field generates a template, indoor The intelligent vehicle task and target map recognition information are determined; the speed potential field is: the speed distribution considered in path planning (for example: the speed limit in a certain area is 40km / h), which affects the vehicle's driving speed in different areas. The speed potential field is determined based on the speed potential field generation template and the target map recognition information; the obstacle repulsion potential field is: a virtual force field that simulates the repulsion effect of obstacles on vehicles. The obstacle repulsion potential field is determined based on the obstacle repulsion potential field generation template and the target map recognition information; the reference line potential field generation template, the speed potential field generation template and the obstacle repulsion potential field generation template are all pre-configured manually;

[0083] Step 3: Real-time scanning is performed by using a laser radar preset on the indoor smart car to generate candidate trajectories; wherein the candidate trajectories are obtained based on the real-time scanning results based on the dynamic obstacle avoidance trajectory planning technology;

[0084] Step 4: Update the memory list in real time according to the trajectory evaluation results of the candidate trajectories, and control the smart car in the room to drive and continue scanning according to the updated results. Among them, when evaluating the trajectory evaluation results, a comprehensive evaluation is made through weighted evaluation of safety loss and stability, deviation from the reference line index and dynamic obstacle distance evaluation index, with the goal of selecting the optimal trajectory to achieve real-time avoidance of dynamic obstacles by the smart car; the memory list is: a data structure in which the smart car stores the latest planned best candidate trajectory.

[0085] The working principle and beneficial effects of the above technical solution are:

[0086] The present invention converts the recognized indoor map recognition information into the Frenet space coordinate system to obtain the target map recognition information; based on the improved artificial potential field algorithm and according to the target map recognition information, the reference line potential field, the velocity potential field and the obstacle repulsion potential field are introduced; the laser radar is controlled to scan in real time, the optimal trajectory with the best evaluation is screened and the memory list is updated in real time, and after the update, the indoor smart car drives based on the optimal trajectory and then continues to scan to perform local trajectory planning, so that the trajectory selection is highly efficient and the driving stability is higher.

[0087] In one embodiment, obtaining indoor map identification information includes:

[0088] Based on the preset lidar and IMU sensors on the indoor smart car, indoor map recognition information is obtained.

[0089] The working principle and beneficial effects of the above technical solution are:

[0090] The present invention introduces laser radar and IMU sensors to obtain indoor map recognition information, thereby improving the indoor map recognition accuracy.

[0091] In one embodiment, step 2: converting the indoor map recognition information into the target map recognition information of the Frenet space coordinate system, and introducing the reference line potential field, the velocity potential field and the obstacle repulsion potential field, includes:

[0092] Obtaining vehicle parameters; wherein the vehicle parameters include: vehicle size, weight, maximum speed, acceleration and steering characteristics, etc.;

[0093] According to the vehicle parameters, the indoor map recognition information in the Cartesian coordinate system is converted into the target map recognition information in the Frenet space coordinate system;

[0094] According to the vehicle parameters, the artificial potential field algorithm is improved, and the improved artificial potential field algorithm includes: adding reference line potential field, velocity potential field and obstacle repulsion potential field.

[0095] The working principle and beneficial effects of the above technical solution are:

[0096] The present invention introduces vehicle parameters to perform coordinate transformation and artificial potential field improvement, thereby improving the accuracy of subsequent extracted evaluation indicators.

[0097] In one embodiment, step 3: performing real-time scanning by a laser radar preset on the indoor smart car to generate candidate trajectories includes:

[0098] Real-time scanning is performed through the preset laser radar on the indoor smart car, and multiple candidate trajectories are planned based on dynamic obstacle avoidance trajectory planning technology.

[0099] The working principle and beneficial effects of the above technical solution are:

[0100] The present invention plans candidate trajectories based on real-time scanning results and dynamic obstacle avoidance trajectory planning technology, thereby improving the suitability of candidate trajectory generation.

[0101] In one embodiment, a laser radar preset on an indoor smart car performs real-time scanning, and plans multiple candidate trajectories based on a dynamic obstacle avoidance trajectory planning technology, including:

[0102] According to the real-time scanning results, try to obtain a suspicious point cloud set; wherein the suspicious point cloud set is: a fuzzy point cloud or a sparse point cloud;

[0103] If the acquisition attempt is successful, the target identification object set is determined; wherein the target identification object set is: a set of possible identification objects of the suspicious point cloud;

[0104] Obtain scanning parameters of the suspicious point cloud set; wherein the scanning parameters are: setting parameters of the scanning radar of the suspicious point cloud set, such as: scanning distance, scanning frequency, etc.;

[0105] According to the scanning parameters, a scanning point cloud set of the target identification object in the target identification object set is obtained; wherein the scanning point cloud set is: a point cloud data set of the target identification object corresponding to the scanning parameters;

[0106] Perform point cloud matching on the suspicious point cloud set and the scanned point cloud set to determine the matching identification object;

[0107] Obtain an actual point cloud set of the matching identification object corresponding to the scanning parameters;

[0108] Perform dynamic obstacle avoidance and plan candidate trajectories based on the actual point cloud set;

[0109] Wherein, obtaining a scanning point cloud set of an identification object in an identification object set according to the scanning parameters includes:

[0110] Obtain historical scanning parameters of the historical scanning point cloud set of the target object, and if the historical scanning parameters match the scanning parameters, use the corresponding historical scanning point cloud set as the scanning point cloud set;

[0111] and / or,

[0112] A scanning simulation of the target object is performed according to the scanning parameters, and a simulated point cloud set obtained by the scanning simulation is used as a scanning point cloud set.

[0113] The working principle and beneficial effects of the above technical solution are:

[0114] The premise of planning candidate trajectories based on dynamic obstacle avoidance trajectory planning technology is to clearly identify obstacles. However, when the laser radar is scanning, there are problems such as point cloud ambiguity and point cloud sparseness that make it impossible to identify. The present invention obtains a suspicious point cloud set, and at the same time, obtains a set of possible target identification objects in the suspicious point cloud set, determines the scanning point cloud set based on the scanning parameters, performs point cloud matching on the suspicious point cloud set and the scanning point cloud set, and uses the actual point cloud set whose matching identification objects actually correspond to the scanning parameters as subsequent obstacle avoidance planning, removes noise interference from the suspicious point cloud set, makes obstacle avoidance target identification more accurate, and makes subsequent planning more suitable.

[0115] In one embodiment, determining a target identifier set includes:

[0116] Try to obtain a scanned confirmed object within the preset target range of the suspicious point cloud set; wherein the preset target range is manually preset, such as: 10 meters; the scanned confirmed object is: an object that can be clearly identified within the target range of the suspicious point cloud set;

[0117] If the acquisition attempt is successful, a relationship feature set is extracted based on the scanned confirmed object and the suspicious point cloud set, and the relationship features include: the type of the scanned confirmed object and the distance between the scanned confirmed object and the suspicious point cloud set;

[0118] According to the spatial distribution of the identification object set in the space where the indoor smart car is located, a relationship feature set library is constructed; wherein the relationship feature set library includes relationship features between multiple identification objects and other identification objects within the surrounding target range;

[0119] Determine a target identification object set according to the relationship feature set and the relationship feature set library;

[0120] If the acquisition attempt fails, the identification object set of the space where the indoor smart car is located is used as the target identification object set.

[0121] The working principle and beneficial effects of the above technical solution are:

[0122] When the present invention determines the possible identification objects of the suspicious point cloud set, it obtains the scanned confirmed objects within the target range of the suspicious point cloud set, and extracts the relationship feature set between the scanned confirmed objects and the suspicious point cloud set, and matches the relationship feature set with a relationship feature set library constructed based on the spatial distribution of the identification object set in the space where the indoor smart car is located. If there is a match, the target identification object set in the identification object set is screened out to improve the subsequent recognition efficiency. Otherwise, the identification object set is directly used as the target identification set for subsequent matching.

[0123] In one embodiment, step 4: updating the memory list in real time according to the trajectory evaluation results of the candidate trajectories, and controlling the indoor smart car to drive and continue scanning according to the updated results, includes:

[0124] The candidate trajectory is evaluated based on the weighted evaluation of safety loss and comfort, acceleration verification, curvature verification, speed verification, deviation from the reference line index and dynamic obstacle distance evaluation index; wherein, when evaluating based on the above multiple evaluation indicators, the quantization template preset for the corresponding indicator type is used for quantization (preset manually), and all quantized values ​​are summed to obtain the quantized score of the candidate trajectory (i.e., the trajectory evaluation result of the candidate trajectory);

[0125] According to the trajectory evaluation results of the candidate trajectories, an optimal trajectory, a collision trajectory and a suboptimal trajectory are determined;

[0126] Put the optimal trajectory into the memory list, and put the collision trajectory and suboptimal trajectory into the forget list;

[0127] The indoor smart car is controlled to drive and continue scanning according to the optimal trajectory updated in real time according to the memory list.

[0128] The working principle and beneficial effects of the above technical solution are:

[0129] The present invention introduces safety loss and comfort weighted evaluation, acceleration verification, curvature verification, speed verification, deviation from reference line index and dynamic obstacle distance evaluation index to evaluate candidate trajectories, thereby improving the accuracy of the evaluation; according to the trajectory evaluation results of the candidate trajectories, the optimal trajectory is put into a memory list, and the collision trajectory and the suboptimal trajectory are put into a forget list, thereby reducing the amount of calculation, reducing the probability of misselection, and improving the generalization ability.

[0130] The embodiment of the present invention provides a path planning method based on an indoor smart car, further comprising:

[0131] When the smart car in the control room drives and continues scanning according to the update results, the relay scanning data is corrected to obtain the target relay scanning data.

[0132] The working principle and beneficial effects of the above technical solution are:

[0133] When the indoor smart car reaches the terminal trajectory point of the most recently updated optimal trajectory, it needs to scan again at the terminal trajectory point to determine a new local trajectory (for example, the indoor smart car rotates around the terminal trajectory point once). However, the scanning angle based on the terminal trajectory point is single, and the subsequent obstacle recognition is not accurate enough. Therefore, when the indoor smart car moves based on the most recently updated optimal trajectory, it is also controlled to scan. After reaching the end point of the stage, the three-dimensional data at different angles are integrated to improve the scanning accuracy.

[0134] In one embodiment, when the smart car in the control room drives and continues scanning according to the update result, the relay scanning data is corrected to obtain the target relay scanning data, including:

[0135] Obtain the scanning range information of the indoor smart car; the scanning range information is: the area range information that the laser radar carried by the smart car can cover when scanning the environment, including parameters such as distance and angle, for example: the horizontal field of view angle is 98.4°, the vertical field of view angle is 38.4°, and the detection distance is 220 meters;

[0136] Determine the trajectory points in the updated optimal trajectory, and determine the target tangent lines of the trajectory points;

[0137] The scanning range information is visualized according to the scanning range information and the target tangent, and the process coverage area is determined; wherein, when the scanning range information is visualized according to the scanning range information and the target tangent, the track point divides the target tangent into two rays, and the ray direction consistent with the driving direction of the smart car is used as the direction of the laser radar probe. After the direction of the laser radar probe is determined, the process coverage area is determined according to the scanning range information;

[0138] Obtain a target coverage area of ​​a process coverage area during the driving process of the indoor smart car based on the updated optimal trajectory; wherein the target coverage area is the sum of the process coverage areas;

[0139] Determine the end point of the updated optimal trajectory;

[0140] Determine the relay scanning area according to the terminal track point and the scanning range information; wherein the relay scanning area is: a scanning area that is circled around the terminal track point;

[0141] Determine the overlapping area between the target coverage area and the relay scanning area;

[0142] Obtaining scanning point correspondence information of the overlapped area and the relay scanning area; wherein the scanning point correspondence information is: which scanning point in the overlapped area corresponds to which scanning point in the relay scanning area;

[0143] According to the corresponding information of the scanning points, a fusion relationship of the scanning data corresponding to the scanning points is determined; wherein, when determining the fusion relationship, there is a fusion relationship between the three-dimensional data of the scanning points corresponding to the scanning points;

[0144] The scanning data is fused according to the fusion relationship to obtain the target relay scanning data.

[0145] The working principle and beneficial effects of the above technical solution are:

[0146] The present invention determines the probe orientation of the laser radar probe according to the target tangent of the track point in the updated optimal track and the driving orientation of the smart car, determines the process coverage area corresponding to the track point according to the probe orientation and the scanning range information, and when the indoor smart car reaches the terminal track point, determines the area sum of the process coverage areas of all track points as the target coverage area, and at the same time, obtains the relay scanning area of ​​the smart car based on the terminal track point.

[0147] The overlapping area between the target coverage area and the relay scanning area is determined. The scanning points in the overlapping area are scanned at least twice, and the scanning directions are inconsistent. The scanning data of different scanning angles corresponding to the same scanning point are fused into point clouds to obtain the target relay scanning data, which improves the scanning recognition accuracy and makes the subsequent path planning more suitable.

[0148] The embodiment of the present invention provides a path planning method based on an indoor smart car, further comprising:

[0149] Obtaining task evaluation information after the indoor smart car completes the driving task; wherein the task evaluation information is: information on the user's evaluation and feedback on the performance of the indoor smart car after completing the driving task;

[0150] Parse the task evaluation information to obtain the first negative evaluation semantics of the indoor smart car; wherein the first negative evaluation semantics is obtained by semantic parsing of the task evaluation record whose score is lower than the score threshold, such as "the passability is too poor, the threshold stone cannot be passed, the ground track cannot be passed", etc.;

[0151] Obtaining the second negative evaluation semantics of the negative evaluation type preset by the indoor smart car; wherein the negative evaluation type includes: inaccurate obstacle recognition, poor passability, etc.;

[0152] Semantically matching the first negative evaluation semantics with the second negative evaluation semantics to determine the target negative evaluation type and its corresponding third negative evaluation semantics that meet the semantic matching requirements; wherein the semantic matching requirement means that the semantic matching similarity is greater than or equal to a preset semantic matching similarity threshold;

[0153] When the target negative evaluation type is inaccurate obstacle recognition, the obstacle type is determined according to the third negative evaluation semantics; wherein the obstacle type is: an obstacle that is inaccurately recognized by the indoor smart car, such as a data cable that falls on the ground;

[0154] Setting scanning scenes of target obstacles of different obstacle types and multiple placement states for the preset target detection model of the indoor smart car to learn;

[0155] When the target negative evaluation type is poor passability, a target pass height set is determined according to the third negative evaluation semantics; wherein the target pass height set is: a set of heights that the indoor smart car cannot pass when the target negative evaluation type is poor passability and is determined by parsing the third negative evaluation semantics;

[0156] Obtaining a theoretical passing height of an indoor smart car; wherein the theoretical passing height is: a passing height set when the indoor smart car is manufactured;

[0157] Calculate the number of occurrences of target passing heights that are lower than the theoretical passing height;

[0158] If the number of occurrences is greater than or equal to the number of occurrences threshold, the third negative evaluation semantics corresponding to the target passing height lower than the theoretical passing height is used as the fourth negative evaluation semantics; wherein the number of occurrences threshold is one thousandth of the number of indoor smart cars produced and applied;

[0159] Based on the preset passability impact basis extraction template and the fourth negative evaluation semantics, the passability impact basis is determined; wherein the passability impact basis extraction template is a template for extracting the passability impact basis of the indoor smart car according to the fourth negative evaluation semantics, and includes multiple preset passability impact basis description semantics, such as: the climbing ability becomes weaker when the battery is insufficient, the wheel slips when climbing, and it is easy to roll over when climbing, etc.;

[0160] According to the passability impact basis, the passability of the indoor smart car is optimized. When the passability of the indoor smart car is optimized according to the passability impact basis, the subsequent passability of the indoor smart car is optimized according to the preset historical experience strategy based on the determined passability impact basis. For example, when the passability impact basis is that the climbing ability weakens when the power is insufficient, the historical experience strategy is: when the power is lower than the set power, forced power conservation is performed.

[0161] The working principle and beneficial effects of the above technical solution are:

[0162] The present invention introduces task evaluation information after the indoor smart car driving task is completed, determines the target negative evaluation type and its corresponding third negative evaluation semantics; when the target negative evaluation type is inaccurate obstacle recognition, determines the obstacle type according to the third negative evaluation semantics, sets scanning scenes of multiple placement states of target obstacles of the obstacle type for learning by a preset target detection model of the indoor smart car, and improves the obstacle recognition accuracy in a targeted manner based on user evaluation; when the target negative evaluation type is poor passability, determines a set of heights that the indoor smart car cannot pass according to the third negative evaluation semantics, and at the same time, obtains the theoretical pass height of the indoor smart car, and when the number of occurrences of the target pass height that is lower than the theoretical pass height is greater than or equal to the number of occurrence thresholds, the third negative evaluation semantics corresponding to the target pass height that is lower than the theoretical pass height is used as the fourth negative evaluation semantics; introduces a passability influence basis extraction template to extract the passability influence basis in the fourth negative evaluation semantics, and obtains the passability improvement strategy in a targeted manner according to the passability influence basis to optimize the passability of the indoor smart car, thereby avoiding the omission of passability influence factors in the improvement process caused by incomplete consideration of the improvement personnel, and improving the improvement efficiency.

[0163] The embodiment of the present invention provides a path planning system based on an indoor smart car, such as Figure 2 As shown, including:

[0164] Indoor map identification information acquisition subsystem 1, used to acquire indoor map identification information;

[0165] Coordinate conversion subsystem 2, used to convert indoor map recognition information into target map recognition information of Frenet space coordinate system, and introduce reference line potential field, velocity potential field and obstacle repulsion potential field;

[0166] The candidate trajectory generation subsystem 3 is used to generate candidate trajectories by real-time scanning using a laser radar preset on the indoor smart car;

[0167] The continuous scanning subsystem 4 is used to update the memory list in real time according to the trajectory evaluation results of the candidate trajectories, and control the indoor smart car to drive and continue scanning according to the updated results.

[0168] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A path planning method based on an indoor intelligent vehicle, characterized in that: include: Step 1: Obtain indoor map identification information; Step 2: Convert the indoor map recognition information into the target map recognition information of the Frenet space coordinate system, and introduce the reference line potential field, velocity potential field and obstacle repulsion potential field; Step 3: Use the laser radar preset on the indoor smart car to perform real-time scanning to generate candidate trajectories; Step 4: The memory list is updated in real time according to the trajectory evaluation results of the candidate trajectories, and the smart car in the control room drives and continues scanning according to the updated results.

2. A path planning method based on an indoor intelligent vehicle as claimed in claim 1, characterized in that: Step 1: Obtain indoor map identification information, including: Based on the preset lidar and IMU sensors on the indoor smart car, indoor map recognition information is obtained.

3. A path planning method based on an indoor intelligent vehicle as claimed in claim 1, characterized in that: Step 2: Convert the indoor map recognition information into the target map recognition information of the Frenet space coordinate system, and introduce the reference line potential field, velocity potential field and obstacle repulsion potential field, including: Get vehicle parameters; According to the vehicle parameters, the indoor map recognition information in the Cartesian coordinate system is converted into the target map recognition information in the Frenet space coordinate system; According to the vehicle parameters, the artificial potential field algorithm is improved, and the improved artificial potential field algorithm includes: adding reference line potential field, velocity potential field and obstacle repulsion potential field.

4. A path planning method based on an indoor intelligent vehicle as claimed in claim 1, characterized in that: Step 3: Use the laser radar preset on the indoor smart car to perform real-time scanning to generate candidate trajectories, including: Real-time scanning is performed through the preset laser radar on the indoor smart car, and multiple candidate trajectories are planned based on dynamic obstacle avoidance trajectory planning technology.

5. A path planning method based on an indoor intelligent vehicle as claimed in claim 4, characterized in that: The laser radar preset on the indoor smart car performs real-time scanning and plans multiple candidate trajectories based on dynamic obstacle avoidance trajectory planning technology, including: According to the real-time scanning results, try to obtain the suspicious point cloud set; If the acquisition attempt is successful, the target identification set is determined; Obtain scanning parameters of the suspicious point cloud set; According to the scanning parameters, a scanning point cloud set of the target identification object in the target identification object set is obtained; Perform point cloud matching on the suspicious point cloud set and the scanned point cloud set to determine the matching identification object; Obtain an actual point cloud set of the matching identification object corresponding to the scanning parameters; Perform dynamic obstacle avoidance and plan candidate trajectories based on the actual point cloud set; Wherein, obtaining a scanning point cloud set of an identification object in an identification object set according to the scanning parameters includes: Obtain historical scanning parameters of the historical scanning point cloud set of the target object, and if the historical scanning parameters match the scanning parameters, use the corresponding historical scanning point cloud set as the scanning point cloud set; and / or, A scanning simulation of the target object is performed according to the scanning parameters, and a simulated point cloud set obtained by the scanning simulation is used as a scanning point cloud set.

6. A path planning method based on an indoor intelligent vehicle as claimed in claim 5, characterized in that: Determine the target identifier set, including: Try to obtain scan confirmation objects within the preset target range of the suspicious point cloud set; If the acquisition attempt is successful, a relationship feature set is extracted based on the scanned confirmed object and the suspicious point cloud set, and the relationship features include: the type of the scanned confirmed object and the distance between the scanned confirmed object and the suspicious point cloud set; According to the spatial distribution of the identification object set in the space where the indoor smart car is located, a relational feature set library is constructed; Determine a target identification object set according to the relationship feature set and the relationship feature set library; If the acquisition attempt fails, the identification object set of the space where the indoor smart car is located is used as the target identification object set.

7. A path planning method based on an indoor intelligent vehicle as claimed in claim 1, characterized in that: Step 4: Update the memory list in real time according to the trajectory evaluation results of the candidate trajectories, and control the smart car in the room to drive and continue scanning according to the updated results, including: Candidate trajectories are evaluated based on weighted assessment of safety loss and comfort, acceleration check, curvature check, speed check, deviation from reference line index, and dynamic obstacle distance evaluation index; According to the trajectory evaluation results of the candidate trajectories, an optimal trajectory, a collision trajectory and a suboptimal trajectory are determined; Put the optimal trajectory into the memory list, and put the collision trajectory and suboptimal trajectory into the forget list; The indoor smart car is controlled to drive and continue scanning according to the optimal trajectory updated in real time according to the memory list.

8. A path planning method based on an indoor intelligent vehicle as claimed in claim 1, characterized in that: Also includes: When the smart car in the control room drives and continues scanning according to the update results, the relay scanning data is corrected to obtain the target relay scanning data.

9. A path planning method based on an indoor intelligent vehicle as claimed in claim 8, characterized in that: When the smart car in the control room drives and continues scanning according to the update results, the relay scanning data is corrected to obtain the target relay scanning data, including: Get the scanning range information of the indoor smart car; Determine the trajectory points in the updated optimal trajectory, and determine the target tangent of the trajectory points; Visualize the scanning range information and determine the process coverage area based on the scanning range information and target tangent; Obtain the target coverage area of ​​the process coverage area during the driving process of the indoor smart car based on the updated optimal trajectory; Determine the end point of the updated optimal trajectory; Determine the relay scanning area based on the endpoint trajectory point and scanning range information; Determine the overlapping area between the target coverage area and the relay scanning area; Obtain the corresponding information of the scanning points in the overlapped area and the relay scanning area; According to the corresponding information of the scanning points, the fusion relationship of the scanning data corresponding to the scanning points is determined; The scanning data is fused according to the fusion relationship to obtain the target relay scanning data.

10. A path planning system based on an indoor intelligent vehicle, characterized in that: include: Indoor map identification information acquisition subsystem, used to acquire indoor map identification information; The coordinate conversion subsystem is used to convert the indoor map recognition information into the target map recognition information of the Frenet space coordinate system, and introduce the reference line potential field, velocity potential field and obstacle repulsion potential field; The candidate trajectory generation subsystem is used to generate candidate trajectories by real-time scanning using a laser radar preset on the indoor smart car; The continuous scanning subsystem is used to update the memory list in real time according to the trajectory evaluation results of the candidate trajectories, and control the indoor smart car to drive and continue scanning according to the updated results.

Images