High-speed automatic driving decision-making system and method based on frequency division triggering and vehicle

Abstract

The invention provides a high-speed automatic driving decision-making system and method based on frequency division triggering and a vehicle, and the system comprises an environment frame module which is used for processing collected external data, and converting the external data into first data; the decision module is used for generating a reasonable scene decision according to the first data; the scene decision arbitration module is used for arbitrating whether the scene decision is feasible or not and generating an arbitration result; the scene state module is used for determining a target scene according to the scene decision and the arbitration result, and switching the current scene to the target scene; the scene corresponding planner is used for selecting a proper planner to perform trajectory planning according to the target scene; triggering conditions of the decision module and the scene decision arbitration module are different, and when the triggering conditions are met, the corresponding module is triggered; the trigger condition is metered by the planning period and the frame number of arbitration failure. The technical problem that the planning module and the decision-making module adopt the same planning cycle, and the stability of the planning track is possibly reduced is solved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a high-speed autonomous driving decision-making system based on frequency division triggering, a high-speed autonomous driving decision-making method based on frequency division triggering, and a vehicle. Background Art

[0002] In most existing autonomous driving systems, the planning module and the decision-making module are operated simultaneously in the same planning cycle. That is, in a planning cycle, the decision-making module first provides the corresponding external information processing results, including but not limited to the driver's driving style, environmental information, scene information, etc. The planning module then makes the corresponding path planning. The advantage of this is that each planning cycle can accurately obtain the corresponding environmental information and make timely judgments. This triggering method is more in line with human driving habits in cities, that is, real-time acquisition of environmental information and corresponding processing, and then converted into the actions of the accelerator, brake, steering wheel, and turn signal.

[0003] However, in the actual implementation process, there is such a problem: the interaction between the decision module and the planning module in the existing decision system adopts a serial mode and adopts time-triggered control, that is, maintaining the agreed release frequency. In some special cases, such as other vehicles driving on the line, the decision of the self-vehicle may fluctuate, resulting in reduced stability of the planned trajectory. Summary of the invention

[0004] The present invention solves the technical problem that the stability of the planning trajectory may be reduced when a planning module and a decision module adopt the same planning cycle.

[0005] To solve the above problems, the present invention provides a high-speed automatic driving decision system based on frequency division triggering, comprising: an environment frame module, the environment frame module is used to process the collected external data and convert the external data into first data; a decision module, the decision module generates a reasonable scene decision according to the first data; a scene decision arbitration module, the scene decision arbitration module is used to arbitrate whether the scene decision is feasible and generate an arbitration result, wherein the arbitration result includes: arbitration passed and arbitration failed; a scene state module, the scene state module is used to determine the target scene according to the scene decision and the arbitration result, and switch the current scene to the target scene; a scene corresponding planner, the scene corresponding planner is used to select a suitable planner for trajectory planning according to the target scene; wherein the trigger conditions of the decision module, the scene decision arbitration module and the scene corresponding planner are different, and when the trigger conditions are met, the corresponding modules are triggered; the trigger conditions of the decision module and the scene decision arbitration module are measured by the planning cycle and the number of frames of arbitration failure; the trigger condition of the scene corresponding planner is every 1 planning cycle.

[0006] Compared with the prior art, the technical effect achieved by adopting this technical solution is as follows: the triggering conditions of the decision module, the scene decision arbitration module and the scene corresponding planner are all different. The scene corresponding planner is triggered once every 1 planning cycle, and the triggering conditions of the decision module and the scene decision arbitration module are measured by the planning cycle and the number of frames in which arbitration fails. In this application, different triggering conditions are adopted for the decision module, the scene decision arbitration module and the scene corresponding planner to reduce the computing pressure on the vehicle computing platform; at the same time, by giving different frequencies to different modules, starting from the user's driving behavior, low-frequency decision-making and high-frequency planning can effectively increase stability while ensuring performance. At the same time, in order to increase the response to special situations, an event triggering method is introduced (that is, triggered by measuring the number of frames in which arbitration fails). That is to say, in the specified dangerous situation, the scene decision re-selection is directly entered to reduce the harm caused to the vehicle by the irregular driving of other vehicles.

[0007] In one example of the present invention, the decision module includes: a scenario decision module, which is used to generate a reasonable first decision based on first data when the vehicle does not comply with the lane keeping rules; wherein the lane keeping rules are rules that allow the vehicle to continue to perform lane keeping; the first decision includes: lever change, free lane change, ramp merging and ramp merging; the triggering conditions of the scenario decision module are divided into a first time triggering condition and a first event triggering condition; the first time triggering condition is: every a planning cycles; the first event triggering condition is: the arbitration result is arbitration failure.

[0008] Compared with the prior art, the technical effect achieved by adopting this technical solution is as follows: the triggering mode of the scene decision module is divided into the first time trigger and the first event trigger. When the first time trigger and / or the first event trigger are met, the scene decision module is triggered. The first time trigger is triggered once every a planning cycle, where a is an integer greater than 1; the event trigger is the failure of scene decision arbitration. The first time trigger frequency is set to trigger once in multiple planning cycles, that is, low-frequency triggering, because in high-speed scenarios, high speeds bring higher dangers to scenes other than straight driving; lane changes and borrowing at low speeds can be carried out more easily and frequently, but at high speeds, a slight movement of the steering wheel will cause terrible consequences; therefore, at high speeds, the stability of the scene is more important, and the switching frequency does not need to be so high. The first event trigger frequency is the failure of scene decision arbitration, which means that the current scene switches to the target scene. It may be due to external static environment, dynamic environment, human factors, etc., and the scene switch is dangerous or canceled manually, and the decision needs to be recalculated.

[0009] In one embodiment of the present invention, the decision module further includes: a lane keeping decision module, which is used to generate a reasonable second decision based on the first data when the vehicle complies with the lane keeping rule; wherein the second decision includes: cruising and following the vehicle; the triggering condition of the lane keeping decision module is divided into a second time triggering condition and a second event triggering condition; the second time triggering condition is: every 1 planning cycle; the second event triggering condition is: continuous The frame arbitration result is arbitration failure.

[0010] Compared with the existing technology, the technical effect achieved by adopting this technical solution is as follows: the scene decision module mainly involves judgments in multi-lane and congested scenes. After the environmental frame module ends, the lane keeping decision module and the scene decision module are synchronously entered. At high speeds, lane keeping is the basis of all other scenes, and it is also the most important and most commonly used scene, occupying most of the time during normal driving. Therefore, the second time trigger condition is triggered once every 1 planning cycle, that is, high-frequency triggering. Because of the right of way of the self-lane, it can also be regarded as a bottom-up module. The second event trigger frequency is continuous Failure of frame scene decision arbitration means that the current scene switches to the target scene, and the scene switching occurs multiple times in a row, which is dangerous or manually canceled, and the decision needs to be recalculated. When the second time trigger and / or the second event trigger is met, the lane keeping decision module is triggered.

[0011] In one embodiment of the present invention, the triggering conditions of the scenario decision arbitration module are divided into a third time triggering condition and a third event triggering condition; the third time triggering condition is: every planning cycle; the third event triggering condition is: continuous The frame arbitration result is arbitration failure.

[0012] Compared with the existing technology, the technical effect achieved by adopting this technical solution is: after obtaining the corresponding scenario decision (including the first decision and / or the second decision), it is necessary to arbitrate whether the scenario decision is feasible. When the third time trigger and / or the third event trigger are met, the scenario decision arbitration module is triggered. The third time trigger condition is triggered once every 1 planning cycle; the third event trigger frequency is continuous Frame scene decision arbitration failure means that scene switching occurs multiple times in a row, which is dangerous or manually canceled, and the scene decision arbitration module needs to be recalculated.

[0013] In one example of the present invention, the scenario decision arbitration module includes: a system definition unit, the system definition unit includes: trigger conditions, upstream information judgment, environmental semantic judgment after preliminary analysis, etc.; a scenario judgment unit, the scenario judgment unit is used to make a primary judgment on whether the scenario decision is feasible; a redundant judgment unit, the redundant judgment unit is used to make a secondary judgment on whether the scenario decision is feasible; wherein the system definition unit, the scenario judgment unit and the redundant judgment unit are performed linearly.

[0014] Compared with the existing technology, the technical effect achieved by adopting this technical solution is as follows: the three sub-units of the system definition unit, the scene judgment unit and the redundant judgment unit are performed linearly. The scene judgment unit refers to the in-depth judgment of the scene decision (mainly the lane change scene decision). It is necessary to make a more in-depth and detailed analysis of the vehicle field where the vehicle is located based on the various environmental information obtained, indicate the feasibility and danger of the scene decision, and make a judgment once; set a redundant judgment unit, which only uses the unit of mathematical methods to make a second judgment on the scene decision, to ensure the feasibility and safety of the actual implementation of the scene decision.

[0015] In one example of the present invention, the scene state module includes: a scene switching state machine, which is used to determine the target scene according to the scene decision when the arbitration result is arbitration passed; a scene state machine, which is used to execute the step of switching the current scene to the target scene after determining the target scene.

[0016] Compared with the existing technology, the technical effect achieved by adopting this technical solution is: after the scene switching state machine passes, it will enter the internal state machine of a single scene, and the scene decision will be locked until it fails or is completed, and then the scene switching state machine will be used again to return to the lane keeping scene and perform the next possible scene switching action. When the target scene has been determined, switching from the current scene to the target scene is a gradual process.

[0017] In one example of the present invention, the first data includes: lane line information, obstacle information, prediction information, positioning information, map information, control-by-wire chassis information, other vehicle driving style and other vehicle driving intention.

[0018] Compared with the existing technology, the technical effect achieved by adopting this technical solution is as follows: the environment frame module is used to convert external data into the data type required by the decision module, including lane line information, obstacle information, prediction information, positioning information, map information, and control-by-wire chassis information. At the same time, on this basis, the corresponding other vehicle reference information, including the driving style and driving intention of other vehicles, is further calculated through the pre-trained model or real-time calculation module.

[0019] On the other hand, an embodiment of the present invention further provides a high-speed autonomous driving decision-making method based on frequency division triggering, which is applied to a high-speed autonomous driving decision-making system based on frequency division triggering as in the above-mentioned embodiment 1, and the high-speed autonomous driving decision-making method includes: collecting external data and converting the external data into first data; generating a reasonable scenario decision based on the first data, wherein the scenario decision includes a first decision and a second decision; arbitrating whether the scenario decision is feasible to generate an arbitration result, and the arbitration results include: arbitration passed and arbitration failed; determining a target scene based on the scenario decision and the arbitration result, and switching the current scene to the target scene; selecting a suitable planner for trajectory planning based on the target scene; wherein, when the arbitration result is arbitration passed, it means that the scenario decision is feasible; when the arbitration result is arbitration failed, it means that the scenario decision is not feasible.

[0020] Compared with the prior art, the technical effect achieved by adopting this technical solution is as follows: the high-speed autonomous driving decision-making method in this embodiment is applied to the high-speed autonomous driving decision-making system based on frequency division trigger as in the first embodiment above, and has all the beneficial effects of the high-speed autonomous driving decision-making system based on frequency division trigger as in the first embodiment above, which will not be repeated here.

[0021] In one example of the present invention, a target scene is determined based on a scene decision and an arbitration result, and the current scene is switched to the target scene, including: when there are multiple first decisions, defining the first decision whose arbitration result is passed as a feasible scene; when there are multiple feasible scenes, determining the target scene based on the priority of the feasible scenes themselves.

[0022] Compared with the prior art, the technical effect achieved by adopting this technical solution is as follows: when the vehicle does not comply with the lane keeping rules, after the scenario decision module finishes generating the first decision, if there are multiple first decisions, arbitration is made for the multiple first decisions to generate multiple arbitration results, and the arbitration results correspond to the first decisions one by one. The first decision whose arbitration result is passed by arbitration is screened out and defined as a feasible scenario. When there are multiple feasible scenarios, that is, at the same time, there may be multiple scenario decisions that are valid. Therefore, it is necessary to prioritize all the feasible scenarios that have passed, and determine the target scenario according to the scenario priority. When the vehicle complies with the lane keeping rules, after the lane keeping decision module finishes generating the second decision, the second decision is directly input into the scenario decision arbitration module to arbitrate the corresponding scenario.

[0023] On the other hand, the present invention also provides a vehicle, including: a high-speed automatic driving decision system based on frequency division triggering.

[0024] Compared with the prior art, the technical effect achieved by adopting this technical solution is as follows: the vehicle in this embodiment has all the beneficial effects of the high-speed automatic driving decision-making system based on frequency division triggering as in any embodiment of the present invention, which will not be repeated here.

[0025] After adopting the technical solution of the present invention, the following technical effects can be achieved: (1) Different trigger conditions are adopted for the decision module, scenario decision arbitration module and scenario corresponding planner to reduce the computing pressure on the vehicle computing platform. At the same time, by giving different frequencies to different modules, starting from the user's driving behavior, low-frequency decision-making and high-frequency planning can be used to effectively increase stability while ensuring performance; (2) In order to increase the ability to respond to special situations, an event trigger method is introduced (that is, triggered by measuring the number of frames that fail arbitration). In other words, in specified dangerous situations, the system directly enters the re-selection of scene decisions to reduce the harm caused to the vehicle by the irregular driving of other vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic diagram of the structure of a high-speed automatic driving decision system based on frequency division triggering provided in the first embodiment of the present invention; Figure 2 A flowchart of a high-speed autonomous driving decision-making method based on frequency division triggering provided in Embodiment 2 of the present invention.

[0027] Description of reference numerals: 100, environmental frame module; 200, decision module; 210, scene decision module; 220, lane keeping decision module; 300, scene decision arbitration module; 400, scene state module; 410, scene switching state machine; 420, scene state machine; 500, scene corresponding planner. DETAILED DESCRIPTION

[0028] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] [Example 1] See also Figure 1The present invention provides a high-speed automatic driving decision system based on frequency division triggering, comprising: an environment frame module 100, a decision module 200, a scene decision arbitration module 300, a scene state module 400 and a scene corresponding planner 500, wherein the environment frame module 100 is used to process the collected external data and convert the external data into first data; the decision module 200 generates a reasonable scene decision according to the first data; the scene decision arbitration module 300 is used to arbitrate whether the scene decision is feasible and generate an arbitration result, wherein the arbitration result includes: arbitration passed and arbitration failed; the scene state module 400 is used to determine the target scene according to the scene decision and the arbitration result, and switch the current scene to the target scene; the scene corresponding planner 500 is used to select a suitable planner for trajectory planning according to the target scene; wherein the trigger conditions of the decision module 200, the scene decision arbitration module 300 and the scene corresponding planner 500 are different, and when the trigger conditions are met, the corresponding modules are triggered; the trigger conditions of the decision module 200 and the scene decision arbitration module 300 are measured by the planning cycle and the number of frames of arbitration failure; the trigger condition of the scene corresponding planner 500 is every 1 planning cycle.

[0030] In a specific embodiment, in the prior art, in each planning cycle, the first data collected in each planning cycle is first processed by the decision module 200, and then the corresponding path planning is made by the planning module. If all information is reprocessed in each planning cycle, it will cause a huge computing burden on the computing power platform, and it may even be unable to interact with other modules according to the agreed cycle, resulting in serious problems such as information loss and delay, affecting the performance of the autonomous driving vehicle. Therefore, this application uses a method of giving different frequencies to different modules, starting from the user's driving behavior, low-frequency decision-making and high-frequency planning, while ensuring performance, reducing the computing pressure on the vehicle-mounted computing power platform.

[0031] Specifically, external data is collected through laser radar and / or camera, and the environment frame module 100 serves as an interface between external data and internal data, and is mainly responsible for processing the external data and converting it into the data type (that is, the first data) required by the decision module 200; after the environment frame module 100 ends, it enters the decision module 200, and the decision module 200 includes a scene decision module 210 and a lane keeping decision module 220. Reasonable scene decisions are generated through the decision module 200. The number of scene decisions is at least one, and the scene decisions include but are not limited to: lever change decision, free lane change decision, ramp entry decision, ramp exit decision, cruise decision and following decision.

[0032] After obtaining the corresponding scenario decision, it is necessary to arbitrate whether the scenario decision is feasible. The scenario decision arbitration module 300 determines whether the scenario decision is feasible by generating an arbitration result. When the arbitration result is that the arbitration is passed, it means that the scenario decision is feasible; when the arbitration result is that the arbitration fails, it means that the scenario decision is not feasible. The scene status module 400 is used to determine the target scene based on the scenario decision and the arbitration result, and switch the current scene to the target scene. Switching from the current scene to the target scene is a step-by-step process. Since the planning end points required for different target scenes are different, the scene corresponding planner 500 is set mainly based on the final goal, and the appropriate planner is selected for trajectory planning to ensure the normal operation of the function.

[0033] Among them, the triggering conditions of the decision module 200, the scene decision arbitration module 300 and the scene corresponding planner 500 are all different. The scene corresponding planner 500 is triggered once every 1 planning cycle. The triggering conditions of the decision module 200 and the scene decision arbitration module 300 are measured by the planning cycle and the number of frames that failed arbitration. In this application, different triggering conditions are adopted for the decision module 200, the scene decision arbitration module 300 and the scene corresponding planner 500 to reduce the computing pressure on the vehicle computing platform; at the same time, by giving different frequencies to different modules, starting from the user's driving behavior, low-frequency decision-making and high-frequency planning can effectively increase stability while ensuring performance. At the same time, in order to increase the response to special situations, an event triggering method is introduced (that is, triggered by the number of frames that failed arbitration). That is to say, in the specified dangerous situation, the scene decision re-selection is directly entered to reduce the harm caused to the vehicle by the irregular driving of other vehicles.

[0034] Furthermore, the decision module 200 includes: a scenario decision module 210, which is used to generate a reasonable first decision based on the first data when the vehicle does not comply with the lane keeping rules; wherein the lane keeping rules are rules that allow the vehicle to continue to perform lane keeping; the first decision includes: lever change, free lane change, ramp merging and ramp merging; the triggering conditions of the scenario decision module 210 are divided into a first time triggering condition and a first event triggering condition; the first time triggering condition is: every a planning cycles; the first event triggering condition is: the arbitration result is arbitration failure.

[0035] Specifically, when the vehicle does not comply with the lane keeping rules, it means that the vehicle cannot maintain the current lane and continue to operate: for example, when entering or exiting ramps, the road ahead disappears, or when other vehicles affect the driving of the vehicle (such as the vehicle ahead is much slower than the normal driving speed, and the other vehicle has a strong intention to cut in, etc.), the normal driving of the vehicle will be affected, and the vehicle must make corresponding strategies to deal with it. The environment frame module 100 ends, and the scene decision module 210 is entered. Because it involves the acquisition of road rights, the scene decision module 210 is also a module for improving the driving experience and avoiding risks to a certain extent. The first decision is a lane change scene decision, including but not limited to: lever change, free lane change, ramp merging and ramp merging. The scene decision module can be regarded as a lane change decision module.

[0036] The triggering mode of the scene decision module 210 is divided into the first time trigger and the first event trigger. When the first time trigger and / or the first event trigger are met, the scene decision module 210 is triggered. The first time trigger is triggered once every a planning cycle, where a is an integer greater than 1; the event trigger is the failure of the scene decision arbitration. The first time trigger frequency is set to trigger once in multiple planning cycles, that is, low-frequency triggering, because in high-speed scenarios, high speeds bring higher dangers to scenes other than straight driving; lane changes and borrowing at low speeds can be carried out more easily and frequently, but at high speeds, slight movements of the steering wheel will cause terrible consequences; therefore, at high speeds, the stability of the scene is more important, and the switching frequency does not need to be so high. The first event trigger frequency is the failure of the scene decision arbitration, which means that the current scene switches to the target scene. It may be due to external static environment, dynamic environment, human factors, etc., the scene switch is dangerous or canceled manually, and the decision needs to be recalculated.

[0037] Preferably, the preferred value of a is 5, which means that the first time trigger is triggered once every 5 planning cycles.

[0038] Among them, event triggering mainly refers to function degradation or even function exit, which is determined by the degree of arbitration failure. Because lane keeping decision is the most basic and relatively safest function, if lane keeping decision arbitration fails, it means that the driving of the vehicle is under great threat. At this time, other functions need to intervene to ensure the safety of the vehicle and drive as normally as possible. If it is a minor impact, such as cutting into the lane at a close distance and lane keeping cannot cope with this situation, the function will be downgraded; if it is a serious impact, such as the sudden braking of the vehicle in front, safety functions such as AEB will be triggered at this time.

[0039] Furthermore, the decision module 200 further includes: a lane keeping decision module 220, which is used to generate a reasonable second decision according to the first data when the vehicle complies with the lane keeping rule; wherein the second decision includes: cruising and following the vehicle; the triggering condition of the lane keeping decision module 220 is divided into a second time triggering condition and a second event triggering condition; the second time triggering condition is: every 1 planning cycle; the second event triggering condition is: continuous The frame arbitration result is arbitration failure.

[0040] Specifically, when the vehicle complies with the lane keeping rule, it means that the vehicle needs to maintain the current lane and continue to operate. The second decision is the lane keeping decision, including but not limited to: cruising and following the vehicle.

[0041] The scene decision module 210 is mainly involved in the judgment of multi-lane and congestion scenes. After the environment frame module 100 ends, the lane keeping decision module 220 and the scene decision module 210 are synchronously entered. At high speed, lane keeping is the basis of all other scenes, and it is also the most important and most commonly used scene, occupying most of the time during normal driving. Therefore, the second time trigger condition is triggered once every 1 planning cycle, that is, high-frequency triggering. Because of the right of way of the self-lane, it can also be regarded as a bottom-up module. The second event trigger frequency is continuous Failure of frame scene decision arbitration means that the current scene switches to the target scene, and the scene switching occurs multiple times in a row, which is dangerous or manually canceled, and the decision needs to be recalculated. When the second time trigger and / or the second event trigger is met, the lane keeping decision module is triggered.

[0042] Preferably, The preferred value of is 3, that is, the second event triggering frequency is 3 consecutive frames of scene decision arbitration failure.

[0043] Furthermore, the triggering conditions of the scenario decision arbitration module 300 are divided into a third time triggering condition and a third event triggering condition; the third time triggering condition is: every planning cycle; the third event triggering condition is: continuous The frame arbitration result is arbitration failure.

[0044] Specifically, after obtaining the corresponding scenario decision (including the first decision and / or the second decision), it is necessary to arbitrate whether the scenario decision is feasible. When the third time trigger and / or the third event trigger are met, the scenario decision arbitration module 300 is triggered. The third time trigger condition is triggered once every 1 planning cycle; the third event trigger frequency is continuous The frame scene decision arbitration failure means that the scene switching occurs repeatedly and is dangerous or manually canceled, and the scene decision arbitration module 300 needs to be recalculated.

[0045] Preferably, The preferred value of is 3, that is, the third event triggering frequency is 3 consecutive frames of scene decision arbitration failure.

[0046] Furthermore, the scenario decision arbitration module 300 includes: a system definition unit, a scenario judgment unit and a redundant judgment unit. The system definition unit includes: trigger conditions, upstream information judgment, environmental semantic judgment after preliminary analysis, etc.; the scenario judgment unit is used to make a primary judgment on whether the scenario decision is feasible; the redundant judgment unit is used to make a secondary judgment on whether the scenario decision is feasible; wherein, the system definition unit, the scenario judgment unit and the redundant judgment unit are performed linearly.

[0047] Specifically, the scene decision arbitration is mainly divided into the following three sub-units: system definition unit, scene judgment unit and redundant judgment unit, and the three sub-units are carried out linearly. The system definition unit, that is, the basic demand definition, includes: trigger conditions, upstream information judgment, environmental semantic judgment after preliminary analysis, etc.; the scene judgment unit refers to the in-depth judgment of the scene decision (mainly the lane change scene decision). It is necessary to make a more in-depth and detailed analysis of the vehicle field where the vehicle is located based on the various types of environmental information obtained, indicate the feasibility and danger of the scene decision, and make a judgment. Because the above two may have discrepancies with the actual scene due to errors in external information, or in-depth mining of information, or limitations on the generalization ability of the model; therefore, it is necessary to set up a redundant judgment unit, which only uses mathematical methods to make a second judgment on the scene decision to ensure the feasibility and safety of the actual implementation of the scene decision.

[0048] Furthermore, the scene state module 400 includes: a scene switching state machine 410 and a scene state machine 420. The scene switching state machine 410 is used to determine the target scene according to the scene decision when the arbitration result is that the arbitration is passed; the scene state machine 420 is used to execute the step of switching the current scene to the target scene after determining the target scene.

[0049] Specifically, the scene switching state machine 410 refers to the switching between scenes (mainly the current scene and the target scene), for example: lane keeping → ramp merging, lane keeping → lever changing lanes. The switching between scenes is relatively simple. As long as the decision itself is safe, that is, the scene decision arbitration is passed, the scene switching can be considered to be established. At the same time, there may be multiple scene decisions that are established; therefore, all scene decisions need to be prioritized. When there are multiple scene decisions that are safe and can provide scene switching at the same time, the target scene is selected according to the priority of the scene decision itself.

[0050] After the scene switching state machine 410 is passed, the scene state machine 420 inside the single scene will be entered. The scene decision will be locked until it fails or is completed, and then the scene switching state machine will be used again to return to the lane keeping scene to perform the next possible scene switching action. When the target scene has been determined, switching from the current scene to the target scene is a gradual process. Preferably, taking the lever change as an example, there are triggering the lever, reminding other vehicles based on regulations that the vehicle is about to change lanes and the original scene continues, starting the lane change, entering the target lane, and the lane change ends. In addition to the main branch, it is also necessary to add a state process of returning to the own lane when the lane change fails, the corresponding second lane change wait after returning to the own lane, and returning to the own lane after canceling the lever to handle additional situations to ensure normal driving and lane changing.

[0051] Furthermore, the first data includes: lane line information, obstacle information, prediction information, positioning information, map information, control-by-wire chassis information, other vehicle driving style and other vehicle driving intention.

[0052] Specifically, the environment frame module 100 is used to convert external data into data types required by the decision module 200, including: lane information, obstacle information, prediction information, positioning information, map information, control-by-wire chassis information, etc. At the same time, on this basis, the corresponding other vehicle reference information, including the driving style and driving intention of the other vehicle, is further calculated through a pre-trained model or a real-time calculation module.

[0053] [Example 2] See also Figure 2 , which is a flow chart of a high-speed automatic driving decision method based on frequency division triggering provided in the second embodiment of the present invention. The high-speed automatic driving decision method is applied to the high-speed automatic driving decision system based on frequency division triggering in the first embodiment above. The high-speed automatic driving decision method includes: Step S10: collecting external data and converting the external data into first data; Step S20: generating a reasonable scenario decision according to the first data, wherein the scenario decision includes a first decision and a second decision; Step S30: Arbitrate whether the scenario decision is feasible to generate an arbitration result, where the arbitration result includes: arbitration passed and arbitration failed; Step S40: determining the target scene according to the scene decision and arbitration results, and switching the current scene to the target scene; Step S50: Select a suitable planner to perform trajectory planning according to the target scene; Among them, when the arbitration result is that the arbitration is passed, it means that the scenario decision is feasible; when the arbitration result is that the arbitration fails, it means that the scenario decision is not feasible.

[0054] In a specific embodiment, the high-speed autonomous driving decision-making method in this embodiment is applied to the high-speed autonomous driving decision-making system based on frequency division triggering as in the first embodiment above, and has all the beneficial effects of the high-speed autonomous driving decision-making system based on frequency division triggering as in the first embodiment above, which will not be repeated here.

[0055] Further, step S40 includes: In the case where there are multiple first decisions, defining the arbitration result as the first decision passed by arbitration is a feasible scenario; When there are multiple feasible scenarios, the target scenario is determined according to the priority of the feasible scenarios themselves.

[0056] Specifically, when the vehicle does not comply with the lane keeping rule, after the scenario decision module finishes generating the first decision, if there are multiple first decisions, arbitration is made for the multiple first decisions to generate multiple arbitration results, and the arbitration results correspond to the first decisions one by one. The first decision whose arbitration result is passed is screened out and defined as a feasible scenario. If there are multiple feasible scenarios, that is, at the same time, there may be multiple scenario decisions that are valid. Therefore, it is necessary to prioritize all the feasible scenarios that have passed, and determine the target scenario according to the scenario priority.

[0057] When the vehicle complies with the lane keeping rules, after the lane keeping decision module finishes generating the second decision, the second decision is directly input into the scenario decision arbitration module to arbitrate the corresponding scenario.

[0058] [Example 3] This embodiment also provides a vehicle, including a high-speed automatic driving decision system based on frequency division triggering as in the above-mentioned embodiment 1.

[0059] Specifically, this embodiment can achieve the technical effects corresponding to any technical solution in the above-mentioned embodiment 1, which will not be repeated here.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-speed automatic driving decision system based on frequency division triggering, characterized in that: The high-speed automatic driving decision system comprises: An environment frame module (100), the environment frame module (100) being used to process the collected external data and convert the external data into first data; A decision module (200), wherein the decision module (200) generates a reasonable scenario decision based on the first data; A scenario decision arbitration module (300), the scenario decision arbitration module (300) is used to arbitrate whether the scenario decision is feasible and generate an arbitration result, wherein the arbitration result includes: arbitration passed and arbitration failed; A scene state module (400), the scene state module (400) being used to determine a target scene according to the scene decision and the arbitration result, and to switch a current scene to the target scene; A scene-corresponding planner (500), the scene-corresponding planner (500) being used to select a suitable planner for trajectory planning according to the target scene; The trigger conditions of the decision module (200), the scene decision arbitration module (300) and the scene corresponding planner (500) are all different. When the trigger conditions are met, the corresponding modules are triggered. The trigger conditions of the decision module (200) and the scene decision arbitration module (300) are measured by the planning cycle and the number of frames in which arbitration fails. The trigger condition of the scene corresponding planner (500) is every other planning cycle.

2. The high-speed automatic driving decision system according to claim 1, characterized in that: The decision module (200) comprises: A scenario decision module (210), the scenario decision module (210) being used to generate a reasonable first decision based on the first data when the vehicle does not comply with the lane keeping rule; The lane keeping rule is a rule that allows the vehicle to continue to execute lane keeping; the first decision includes: lever change, free lane change, ramp merging and ramp merging; the triggering condition of the scenario decision module (210) is divided into a first time triggering condition and a first event triggering condition; the first time triggering condition is: every a planning cycles; the first event triggering condition is: the arbitration result is arbitration failure.

3. The high-speed automatic driving decision system according to claim 2, characterized in that: The decision module (200) further comprises: A lane keeping decision module (220), the lane keeping decision module (220) being used to generate a reasonable second decision based on the first data when the vehicle complies with the lane keeping rule; The second decision includes: cruising and following a vehicle; the triggering condition of the lane keeping decision module (220) is divided into a second time triggering condition and a second event triggering condition; the second time triggering condition is: every 1 planning cycle; the second event triggering condition is: continuous The arbitration result described in the frame is arbitration failure.

4. The high-speed automatic driving decision system according to claim 1, characterized in that: The triggering condition of the scenario decision arbitration module (300) is divided into a third time triggering condition and a third event triggering condition; The third time trigger condition is: each of the planning cycles; The third event triggering condition is: The arbitration result described in the frame is arbitration failure.

5. The high-speed automatic driving decision system according to claim 4, characterized in that: The scenario decision arbitration module (300) comprises: System definition unit, the system definition unit includes: trigger conditions, upstream information judgment, environmental semantic judgment after preliminary analysis, etc.; A scene determination unit, the scene determination unit is used to determine whether the scene determination is feasible; A redundant judgment unit, the redundant judgment unit is used to make a secondary judgment on whether the scenario decision is feasible; Wherein, the system definition unit, the scene judgment unit and the redundancy judgment unit are performed linearly.

6. The high-speed automatic driving decision system according to claim 1, characterized in that: The scene state module (400) comprises: A scene switching state machine (410), wherein the scene switching state machine (410) is used to determine the target scene according to the scene decision when the arbitration result is arbitration passed; A scene state machine (420), wherein the scene state machine (420) is used to execute the step of switching the current scene to the target scene after determining the target scene.

7. The high-speed automatic driving decision system according to claim 1, characterized in that: The first data includes: lane line information, obstacle information, prediction information, positioning information, map information, control-by-wire chassis information, other vehicle driving style and other vehicle driving intention.

8. A high-speed automatic driving decision method based on frequency division triggering, the high-speed automatic driving decision method being applied to the high-speed automatic driving decision system based on frequency division triggering as claimed in any one of claims 1 to 7, characterized in that: The high-speed automatic driving decision-making method comprises: Collecting external data and converting the external data into first data; Generate a reasonable scenario decision according to the first data, wherein the scenario decision includes a first decision and a second decision; Arbitrating whether the scenario decision is feasible to generate an arbitration result, wherein the arbitration result includes: arbitration passed and arbitration failed; Determine a target scene according to the scene decision and the arbitration result, and switch the current scene to the target scene; Selecting a suitable planner for trajectory planning according to the target scenario; When the arbitration result is that the arbitration is passed, it means that the scenario decision is feasible; when the arbitration result is that the arbitration fails, it means that the scenario decision is not feasible.

9. The high-speed automatic driving decision-making method according to claim 8, characterized in that: The step of determining a target scene according to the scene decision and the arbitration result, and switching the current scene to the target scene includes: In the case that there are multiple first decisions, defining the arbitration result as the first decision passed by arbitration as a feasible scenario; When there are multiple feasible scenarios, the target scenario is determined according to the priorities of the feasible scenarios themselves.

10. A vehicle, characterized in that: include: A high-speed autonomous driving decision system based on frequency division triggering as described in any one of claims 1 to 7.

Images