WO2023123456A1 - 一种车辆位置的预测方法、装置、车辆及存储介质 - Google Patents

一种车辆位置的预测方法、装置、车辆及存储介质 Download PDF

Info

Publication number
WO2023123456A1
WO2023123456A1 PCT/CN2021/143925 CN2021143925W WO2023123456A1 WO 2023123456 A1 WO2023123456 A1 WO 2023123456A1 CN 2021143925 W CN2021143925 W CN 2021143925W WO 2023123456 A1 WO2023123456 A1 WO 2023123456A1
Authority
WO
WIPO (PCT)
Prior art keywords
lane
target vehicle
predicted
probability
vehicle
Prior art date
Application number
PCT/CN2021/143925
Other languages
English (en)
French (fr)
Inventor
吴易霖
罗元福
Original Assignee
深圳市大疆创新科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳市大疆创新科技有限公司 filed Critical 深圳市大疆创新科技有限公司
Priority to PCT/CN2021/143925 priority Critical patent/WO2023123456A1/zh
Priority to CN202180102020.8A priority patent/CN117897749A/zh
Publication of WO2023123456A1 publication Critical patent/WO2023123456A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
    • G01C21/34—Route searching; Route guidance
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems

Definitions

  • the present application relates to the technical field of automobiles, and in particular to a vehicle position prediction method, device, vehicle and storage medium.
  • one of the objectives of the present application is to provide a vehicle location prediction method, device, vehicle and storage medium, which can accurately predict the intentions of other vehicles on the road.
  • a method for predicting a vehicle position comprising:
  • the lane topology information is used to indicate: whether any lane of each road segment in a plurality of consecutive road segments is directly passable to a lane in the next road segment adjacent to the road segment;
  • a device for predicting a vehicle position including:
  • memory for storing processor-executable program instructions
  • a vehicle including:
  • a computer program product including a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect above are implemented.
  • a machine-readable storage medium where several computer instructions are stored on the machine-readable storage medium, and when the computer instructions are executed, the method described in the above-mentioned first aspect is executed.
  • the present application provides a vehicle location prediction method, device, vehicle, and storage medium, which acquire the road section and lane where the target vehicle is currently located, as well as lane topology information. Since the lane topology information is used to indicate any lane of each road segment in multiple consecutive road segments, whether the lane in the next road segment adjacent to this road segment can be directly passed, it is possible to pass the road segment where the target vehicle is located and the lane topology information To predict the lane where the target vehicle is in the next section of the road, so that it can accurately predict the intention of other vehicles on the road.
  • Fig. 1 is a flowchart of a method for predicting a vehicle position according to an embodiment of the present application.
  • Fig. 2 is a schematic diagram of a lane topology according to an embodiment of the present application.
  • Fig. 3 is a flow chart of a method for predicting a vehicle position according to another embodiment of the present application.
  • Fig. 4 is a flowchart of a method for predicting a vehicle position according to another embodiment of the present application.
  • Fig. 5 is a flow chart of a vehicle position prediction method according to another embodiment of the present application.
  • Fig. 6 is a flowchart of a vehicle position prediction method according to another embodiment of the present application.
  • Fig. 7 is a schematic structural diagram of a device for predicting a vehicle position according to an embodiment of the present application.
  • Fig. 8 is a schematic structural diagram of a vehicle according to an embodiment of the present application.
  • the vehicle collects the environmental information of the environment in which the vehicle is located through the equipped sensors.
  • Sensors can include vision sensors (such as multiple monocular or binocular vision devices), lidar, millimeter wave radar, inertial measurement unit (Inertial Measurement Unit, IMU), global navigation satellite system (such as global positioning system Global Positioning System, GPS ), gyroscope, ultrasonic sensor, electronic compass and barometer and other sensors.
  • the environmental information collected by the sensor may include depth information used to indicate the distance between the object and the sensor or the vehicle, traffic indication information (such as traffic light information, information indicated by signs), lane position information, traffic participant position information, etc. .
  • the vehicle's intentions may include multiple types, such as lane-changing intentions, overtaking intentions, deceleration intentions, and so on.
  • the intention such as using the destination as the expression of the vehicle intention, and for example using the workshop space as the expression of the vehicle intention, and so on.
  • it is often difficult to know the intentions of other vehicles directly. It is necessary to use the environmental information collected by sensors to predict the intentions of other vehicles, and plan the behavior of this car based on the prediction results.
  • the accuracy of the prediction results of other vehicles' intentions will affect the accuracy of the own vehicle's behavior planning.
  • there is no related technology that can better predict the intentions of other vehicles accurately which is a problem that needs to be solved urgently in this field.
  • Moving Object Tracking is based on various preprocessed sensor data to obtain the detection results of continuous moving objects in time, and is an objective expression of the past and current state of moving objects.
  • MOT consists of an outline and a number. The coordinates corresponding to the same moving object at different times are different, but the outline and number should be the same. Contours can be expressed using a coordinate system and a Bounding Box method, or using a coordinate system and a polyhedron/polygon method.
  • pose (Pose) coordinates is also chosen as coordinates in order to better express the current and past behavior of the moving object.
  • the acceleration of a moving object is not appreciable, it is rare to select acceleration as a coordinate.
  • Motion Planning refers to planning the future movement of the vehicle (Ego Vehicle) based on environmental information to achieve the optimum under certain metrics (Metric). Metrics can include minimum energy, minimum acceleration, minimum velocity, no collision, etc.
  • Metrics can include minimum energy, minimum acceleration, minimum velocity, no collision, etc.
  • the output result of motion planning can be trajectory (Trajectory), that is, trajectory planning (Trajectory Planning).
  • the solution space of the trajectory planning is the state space, that is, to solve the state coordinates corresponding to the time parameter t of the vehicle.
  • the state space can include parameters such as position, velocity, acceleration, curvature, etc. When the solution space only considers the position, it is called path planning.
  • Behavior Planning is to plan the future behavior of the vehicle, which is used to guide the motion planning of the vehicle.
  • the behavior planning process includes processing various preprocessed sensor data (Perception) and then outputting behavior (Behavior).
  • the expression method of Behavior can be a reference trajectory, or a semantic behavior (such as forward, backward, left turn, etc.), or a combination of the two.
  • Trajectory Prediction refers to the use of MOT to predict the trajectory of a moving object for several seconds in the future. Generally speaking, it refers to predicting the trajectory of other moving objects other than the vehicle. The result of trajectory prediction can guide the motion planning of the ego vehicle.
  • Behavior Prediction refers to predicting the future behavior of moving objects. Generally speaking, it refers to predicting the behavior of other moving objects other than the vehicle. In order to obtain the trajectory prediction as accurate as possible in each decision-making planning cycle (Planning Cycle) of the vehicle, the hidden behavior of the moving object can be predicted. Such prediction behavior is called behavior prediction.
  • Step 110 Obtain the road section and lane where the target vehicle is currently located
  • Step 120 Acquiring lane topology information, the lane topology information is used to indicate: whether any lane of each road segment in a plurality of continuous road segments, and the lane in the next road segment adjacent to this road segment can pass directly;
  • Step 130 Based on the road section and lane where the target vehicle is currently located, and the lane topology information, predict the lane where the target vehicle is located on the next road section.
  • the target vehicle may be a vehicle around the own vehicle that may affect the behavior planning of the own vehicle. For example, it may be a vehicle in front/rear of the own vehicle, or a vehicle in an adjacent lane. In some embodiments, the target vehicle may be a vehicle within a preset distance range from the host vehicle.
  • the lane information can be identified by using the data collected by the sensors mounted on the vehicle, and the lane topology information can be established in real time based on the lane information.
  • the data collected by the sensor may be image data, and the lane information may include lane position information, lane distribution information, lane number information, and the like.
  • electronic maps such as high-precision maps can be used to obtain road network data, and lane topology information can be pre-established based on the road network data.
  • the road is divided into multiple continuous road sections, and the road surface information of different road sections in the same road may be different.
  • road sections may be divided according to the number of lanes or according to changes in the number of lanes.
  • FIG. 2 the schematic diagram of the lane topology, the circles shown in the figure represent the nodes of the topology graph.
  • the nodes of the topology graph are used to represent the lanes, and the paths between the nodes represent the connectivity between the lanes.
  • road section a includes 3 lanes and road section b includes 4 lanes.
  • road segments can be divided according to whether the lanes are changeable or not. For example, in two consecutive road sections, lanes cannot be changed between the lanes of the previous road section, but the same lane can be changed in the latter road section.
  • road segments may also be divided according to length, for example, every 50 meters in a road is divided into a road segment.
  • the way of dividing road sections may include but not limited to the examples listed above, and may also be a combination of the above examples.
  • a road segment can consist of one or more lanes.
  • road section a includes lanes 4, 1, and 7; road section b includes lanes 5, 2, 8, and 10; road section c includes lanes 6 and 3; road section d includes lanes 12, 15, 16, and 20; road section e Road section f includes lanes 14, 18, 19; road section g includes lanes 21, 22; road section h includes lanes 9, 11.
  • road sections a, b, and c are three consecutive road sections, road sections a, b are two adjacent road sections, road sections b, c are two adjacent road sections, road sections g, h are also adjacent road sections. Adjacent road sections may belong to the same road, or may belong to different roads respectively.
  • the lane topology information is used to indicate whether any lane of each road segment in the plurality of consecutive road segments is directly passable to a lane in a next road segment adjacent to the road segment.
  • the schematic diagram of lane topology shown in FIG. 2 shows that lane 4 in road section a can pass directly with lane 5 in the next adjacent road section b.
  • the lane 17 in the road section e can directly pass through the lanes 18 and 19 in the adjacent next road section f.
  • the lane topology information indicates the passability of lanes in adjacent road sections
  • the lane that the target vehicle can drive to in the next road section can be predicted, and
  • the lane that the target vehicle can travel to in the next section is used as the expression of the target vehicle's intention. Since the lane that the target vehicle may travel to in the next section of the road can be predicted, trajectory prediction and behavior prediction of the target vehicle can be performed more accurately.
  • the vehicle is in lane 5 of road section b as shown in Figure 2, and the target vehicle is currently in lane 20 of road section d, then according to the lane topology information, it can be predicted that the target vehicle can drive to lane 9 of the next road section h . That is to say, it can be predicted that the target vehicle intends to travel to the lane 9 of the road segment h.
  • the behavior of the vehicle can be further planned.
  • a method for predicting the vehicle position proposed by the present application uses the lane to which the target vehicle can drive in the next road section as the expression of the vehicle intention for the first time. Therefore, based on the prediction result of the vehicle intention, the accuracy of the behavior planning made by the vehicle will be correspondingly improved, thereby improving the driving safety.
  • the determination process of the road section and lane where the target vehicle is currently located may include steps as shown in Figure 3:
  • Step 310 Obtain the current location information of the target vehicle
  • Step 320 Obtain the location information of the lane
  • Step 330 Based on the current location information of the target vehicle and the location information of the lane, determine the road section and lane where the target vehicle is currently located.
  • the current location information of the target vehicle may be information indicating the geographic location of the target vehicle, such as the latitude and longitude of the target vehicle, or the coordinate position in the world coordinate system.
  • the vehicle can determine its current location through the positioning module, and obtain the depth information of the target vehicle based on the sensors mounted on the vehicle to determine the distance between the target vehicle and the vehicle. Based on the current position of the target vehicle and the distance to the target vehicle, the current position information of the target vehicle can be determined.
  • the lane location information may include lane distribution information, road section information where the lane is located, and so on. Based on the current location information of the target vehicle and the location information of the lane, the road section and lane where the target vehicle is currently located can be determined.
  • the current road section and lane of the target vehicle may also be determined based on the current road section and lane of the vehicle, the depth information of the target vehicle, and the location information of the lane. Using the depth information of the target vehicle, the distance between the target vehicle and the vehicle can be determined, so that the road section and lane where the target vehicle is located can be calculated according to the current road section and lane of the vehicle.
  • the way to determine the road section and lane where the target vehicle is currently located includes but is not limited to the above two methods. Those skilled in the art can adopt other methods to determine the road section and lane where the target vehicle is currently located according to actual needs, and this application does not make any limit.
  • the lane topology information may indicate whether any lane of a road segment is directly passable to the lane of the next road segment adjacent to the road segment. In some scenarios, if the current lane of the target vehicle can directly pass through two or more lanes in the adjacent next road section, then in the process of predicting the lane of the target vehicle in the next road section, it may include The steps shown in Figure 4:
  • Step 410 Based on the road section and lane where the target vehicle is currently located, and the lane topology information, predict the predicted probability that the target vehicle will drive to any lane of the next road section;
  • Step 420 Predict the lane where the target vehicle is located on the next road section based on the prediction probability.
  • the target vehicle is currently in lane 17 in road segment e, according to the lane topology information, it can be known that lane 17 can directly pass through lane 18 and lane 19 in the adjacent next road segment f.
  • the predicted probability of the target vehicle traveling to any lane of the next road section can be predicted, and then the lane of the target vehicle on the next road section can be predicted based on the predicted probability of any lane.
  • the lane corresponding to the maximum predicted probability may be determined as the lane where the target vehicle is located on the next road section. For example, if the predicted probability of the predicted target vehicle traveling to lane 18 is 80%, and the predicted probability of driving to lane 19 is 20%, then the lane where the predicted target vehicle is located in the next section f is lane 18.
  • a road section may include multiple lanes, and vehicles may change lanes in adjacent lanes or vehicles may drive from a certain lane to another adjacent lane.
  • vehicles may change lanes in adjacent lanes or vehicles may drive from a certain lane to another adjacent lane.
  • the prediction probability of the above-mentioned predicted target vehicle traveling to any lane of the next road section may include steps as shown in Figure 5:
  • Step 510 Based on the current lane, the adjacent lanes of the current lane, and the lane topology information, predict the sequence combination of predicted lanes passed by the target vehicle;
  • Step 520 Predict the predicted probability of the target vehicle passing through the predicted lane sequence combination.
  • the predicted lane sequence combination is at least composed of the lanes on the current road segment and the lanes of the adjacent road segments of the current road segment in sequence.
  • the predicted lane sequence combination may include a lane sequence combination composed of the current lane and the directly passable lanes in the adjacent road segment in sequence.
  • the predicted lane sequence combination can include lane 1, lane 2 and lane 3 are composed of lane sequences in sequence, expressed as lane sequence 1-2-3.
  • the predicted lane sequence combination may also include a lane sequence combination composed of adjacent lanes and directly passable lanes in adjacent road sections in sequence.
  • the adjacent lane may include one or both of the left and right lanes adjacent to the lane where the target vehicle is currently located.
  • the adjacent lanes of the lane 1 where the target vehicle A is currently located include lane 4 and lane 7 .
  • the predicted lane sequence combination may include a lane sequence combination composed of lane 4, lane 5, and lane 6 in sequence, denoted as lane sequence 4-5-6.
  • the predicted lane sequence combination may also include a lane sequence combination composed of lane 7, lane 8, and lane 9 in sequence, expressed as lane sequence 7-8-9.
  • the predicted lane sequence combination may also include a lane sequence combination composed of lane 7, lane 10, and lane 11 in sequence, denoted as lane sequence 7-10-11.
  • the predicted probability of lane sequence 1-2-3 can be predicted, and the predicted probability of lane sequence 4-5-6, lane sequence 7-8-9.
  • the predicted probability of the target vehicle traveling to any lane of the next section, and/or the predicted probability of the target vehicle passing through the combination of predicted lane sequences may be a probability distribution prior manually specified based on empirical values, It can also be obtained by combining historical probabilities of lane sequences.
  • the combination of lane sequences of the target vehicle and its predicted probability can be expressed in many ways. For example, in the above example, an initial probability value can be assigned to each lane sequence through manual designation or historical probability, expressed as: 1 -2-3(P1)/4-5-6(P2)/7-8-9(P3)/7-10-11(P4). Among them, P1-P4 is the predicted probability value of each lane sequence.
  • the predicted probability of the target vehicle traveling to any lane of the next road segment and the predicted probability of the target vehicle passing through the predicted lane sequence combination will also change accordingly.
  • the historical probability can be utilized in time series to improve the accuracy of the forecast probability.
  • the predicted probability can also be predicted based on historical probabilities, then any of the above embodiments may further include the step of: obtaining the historical probability of the target vehicle passing through the historical lane sequence combination determined at the moment before the preset time interval.
  • the preset time interval can be set according to actual needs, such as 1 second or 3 seconds.
  • the historical lane sequence combination is at least composed of the lanes on the historical road segment passed by the target vehicle at the moment before the preset time interval and the lanes of the adjacent road segments of the historical road segment in sequence.
  • the predicted lane sequence combination at the first moment includes: 1-2-3(P1)/4-5-6(P2)/7- 8-9(P3)/7-10-11(P4).
  • the first moment is the moment before the preset time interval of the second moment, and the predicted lane sequence at the first moment is combined into the historical lane sequence at the second moment combination.
  • the predicted probability of the predicted lane sequence combination at the second moment can be predicted.
  • the predicted lane sequence at the second moment may include: 2-3(P5)/5-6(P6)/8-9(P7).
  • the historical probability is determined as the predicted probability of the target vehicle driving to the predicted lane sequence combination. For example, in the above example, the target vehicle A is in lane 1 at the first moment, and the target vehicle A is still in lane 1 after the preset time interval, then the predicted lane sequence combination and its predicted probability at this moment are still the first Predicted lane sequence combinations and their predicted probabilities at each moment.
  • the predicted lane is obtained by assigning the historical probabilities of the historical lane sequence combinations according to the lane topology information Predicted probabilities for sequence combinations.
  • the target vehicle A is in lane 2 at the second moment
  • the predicted lane sequence at the second moment includes: 2-3(P5)/5-6(P6)/8-9(P7).
  • the lane sequence 2-3 is an extension of the lane sequence 1-2-3 at the first moment
  • the predicted probability P5 of the lane sequence 2-3 at the second moment is determined to inherit the predicted probability P1 of the lane sequence 1-2-3 at the first moment .
  • the predicted probability P6 of lane sequence 5-6 inherits the predicted probability P2 of lane sequence 4-5-6
  • the predicted probability P7 of lane sequence 8-9 inherits the predicted probability P3 of lane sequence 7-8-9.
  • the historical probability of the historical lane sequence combination is allocated according to the lane traffic relationship to obtain the predicted probability of the predicted lane sequence combination.
  • the non-one-to-one lane traffic relationship includes that one lane of the historical road section can directly pass through more than two lanes of the current road section. As shown in Figure 2, lane 7 can pass through lane 8 and lane 10 directly. In addition, there are more than two lanes in the historical road section and one lane in the current road section can pass through directly. Both lane 12 and lane 15 can pass through lane 13 directly as shown in Figure 2 .
  • the historical probability can be assigned according to the principle of minimum entropy to obtain the predicted probability of the predicted lane sequence combination.
  • the above-mentioned lane passage relationship is: more than two lanes in the historical section and one lane in the current section can pass directly, and the lane sequence corresponding to one lane in the current section inherits the historical probability of the lane sequence corresponding to the two or more lanes in the historical section.
  • the historical probability of the historical lane sequence combination is combined for prediction. Due to the use of historical probabilities, the accuracy of the behavior prediction of the target vehicle can be greatly improved.
  • the prediction probability in any of the above embodiments may also be predicted based on the motion state information of the target vehicle.
  • the motion state information of the target vehicle can be obtained based on MOT.
  • the predicted probability of the predicted lane sequence combination of the target vehicle can be obtained based on the above-mentioned historical probability and motion state information.
  • the historical probability may be iteratively updated based on the motion state information and the preset update strategy until the change of the obtained probability during the update process is less than the preset change threshold.
  • the obtained probability is called the posterior probability, and the posterior probability is determined as the predicted probability of the combination of predicted lane sequences.
  • the aforementioned preset update strategy may include Bayesian update.
  • any of the above embodiments may further include a step of: performing behavior planning for the vehicle based on the predicted lane of the target vehicle in the next road segment.
  • the motion state information can correct the prediction probability based on the historical probability prediction, which further improves the accuracy of the prediction probability.
  • the accuracy of the prediction probability of the target vehicle is improved, the accuracy of the corresponding behavior planning of the vehicle can also be greatly improved.
  • the target vehicle A is in lane 1 at the first moment
  • the predicted lane sequence combination at the first moment includes: 1-2-3(P1)/4-5-6(P2)/7-8 -9(P3)/7-10-11(P4).
  • the probability of target vehicle A staying in the current lane 1 is 0.6
  • the predicted lane sequence combination at the first moment includes: 1-2-3(0.6)/4-5-6(0.2)/7-8-9(0.1)/7-10-11(0.1).
  • the predicted lane sequence combinations at the second moment include: 2-3(P5)/5-6(P6)/8-9(P7).
  • the predicted probability P5 of lane sequence 2-3 inherits the predicted probability P1 of lane sequence 1-2-3 at the first moment
  • the predicted probability P6 of lane sequence 5-6 inherits the predicted probability P2 of lane sequence 4-5-6
  • the predicted probability P7 of the lane sequence 8-9 inherits the predicted probability P3 of the lane sequence 7-8-9.
  • the combination of predicted lane sequences at the second moment includes: 2-3(0.67)/5-6(0.22)/8-9(0.11).
  • the predicted lane sequence combination at the third moment includes: 2-3(P5)/8-9(P7)/10-11(P8). Since there is no historical lane sequence combination corresponding to the historical road segment in the lane sequence 10-11 at the second moment, it can be known from experience that the probability of right-changing lanes is 0.2, that is, the predicted probability P8 of the manually designated lane sequence 10-11 is 0.2.
  • the lane sequence 2-3 and the lane sequence 8-9 at the third moment are the same as the lane sequence 2-3 and the lane sequence 8-9 at the second moment respectively, so the predicted probability and the lane sequence 2-3 of the third moment
  • the predicted probability of sequence 8-9 directly inherits the predicted probability of lane sequence 2-3 and the predicted probability of lane sequence 8-9 at the second moment.
  • the combination of predicted lane sequences at the third moment includes: 2-3 (0.68)/8-9 (0.11)/10-11 (0.21).
  • the predicted lane sequence combination at the second moment includes 13-14(P13)/17-18(P14)/17-19(P15). Since the lane 12 and lane 15 of the historical section d at the second moment and the lane 13 at the third moment can pass directly, the predicted probability P13 of the lane sequence 13-14 inherits the predicted probability P9 of the lane sequence 15-13-14 and the lane sequence Predicted probability P10 of 12-13-14.
  • the lane sequence 17-18 at the third moment is an extension of the lane sequence 16-17-18 at the second moment; the lane sequence 17-19 at the third moment is an extension of the lane sequence 16-17-19 at the second moment.
  • the predicted probability P14 of the lane sequence 17-18 inherits the predicted probability P11 of the lane sequence 16-17-18; the predicted probability P15 of the lane sequence 17-19 inherits the predicted probability P12 of the lane sequence 16-17-19.
  • the predicted lane sequence combination at the second moment includes 13-14(0.8)/17-18(0.1)/17-19(0.1).
  • a vehicle position prediction method provided by the present application obtains the road section and lane where the target vehicle is currently located, and lane topology information. Since the lane topology information is used to indicate any lane of each road section in multiple continuous road sections, it is related to the Whether the lane in the next road segment adjacent to the road segment is directly passable, that is, the lane topology information indicates the passability of the lane in the adjacent road segment, so based on the lane topology information, the lane that the target vehicle can pass in the next road segment can be predicted.
  • the present application uses the lane to which the target vehicle can drive in the next road section as the expression of the vehicle intention for the first time.
  • the lane topology can be continued in time series, when predicting the current predicted lane sequence combination, the historical probability of the historical lane sequence combination can be combined for prediction. Due to the use of historical probabilities, the accuracy of the behavior prediction of the target vehicle can be greatly improved.
  • the prediction probability of the predicted lane sequence combination when determining the prediction probability of the predicted lane sequence combination, it is also combined with the current motion state information of the target vehicle for prediction.
  • the motion state information can correct the prediction probability based on the historical probability prediction to a certain extent, further improving the prediction performance. probability of accuracy.
  • the accuracy of the prediction probability of the target vehicle is improved, the accuracy of the corresponding behavior planning of the own vehicle can also be greatly improved.
  • the present application also provides a method for predicting the position of the vehicle, including the steps shown in Figure 6:
  • Step 610 Obtain the current location information of the target vehicle and the location information of the lane;
  • Step 620 Based on the current location information of the target vehicle and the location information of the lane, determine the road section and lane where the target vehicle is currently located;
  • Step 630 Acquiring lane topology information, the lane topology information is used to indicate: whether any lane of each road segment in a plurality of consecutive road segments can directly pass through the lane in the next road segment adjacent to the road segment;
  • Step 640 Based on the current lane, the adjacent lanes of the current lane, and the lane topology information, predict the sequence combination of predicted lanes passed by the target vehicle;
  • the predicted lane sequence combination is at least composed of the lanes on the current road segment and the lanes of the adjacent road segments of the current road segment in sequence.
  • Step 650 Obtain the historical probability of the target vehicle passing through the combination of historical lane sequences determined at the moment before the preset time interval;
  • the historical lane sequence combination is at least composed of the lanes on the historical road segment passed by the target vehicle at the moment before the preset time interval and the lanes of the adjacent road segments of the historical road segment in sequence.
  • Step 660 Obtain the motion state information of the target vehicle
  • Step 670 Iteratively update the historical probability based on the motion state information and Bayesian update until the change of the obtained probability is less than a preset change threshold, and the obtained probability is the predicted probability of the predicted lane sequence combination;
  • Step 680 Predict the lane where the target vehicle is located on the next road section based on the prediction probability
  • Step 690 Carry out behavior planning for the vehicle based on the predicted lane of the target vehicle on the next road section.
  • the present application provides a method for predicting the position of a vehicle, which obtains the road section and lane where the target vehicle is currently located, and lane topology information, wherein the target vehicle may be a vehicle around the vehicle that may affect the behavior of the vehicle.
  • the lane topology information is used to indicate whether any lane in each road segment in multiple consecutive road segments can directly pass through the lane in the next road segment adjacent to the road segment. Since the lane topology information indicates the passability of lanes in adjacent road sections, the lanes to which the target vehicle can drive in the next road section can be predicted based on the lane topology information.
  • the present application uses the lane to which the target vehicle can drive in the next road section as the expression of the vehicle intention for the first time.
  • the lane topology can be continued in time series, when predicting the current predicted lane sequence combination, the historical probability of the historical lane sequence combination can be combined for prediction. Due to the use of historical probabilities, the accuracy of target vehicle behavior prediction can be greatly improved.
  • the prediction probability of the predicted lane sequence combination when determining the prediction probability of the predicted lane sequence combination, it is also combined with the current motion state information of the target vehicle for prediction.
  • the motion state information can correct the prediction probability based on the historical probability prediction to a certain extent, further improving the prediction performance. probability of accuracy.
  • the accuracy of the prediction probability of the target vehicle is improved, the accuracy of the corresponding behavior planning of the own vehicle can also be greatly improved.
  • the present application further provides a structural schematic diagram of a vehicle position prediction device as shown in FIG. 7 .
  • the prediction device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and of course may also include hardware required by other services.
  • the processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, so as to realize the vehicle position prediction method described in any of the above embodiments.
  • the present application also provides a schematic structural diagram of a vehicle as shown in FIG. 8 .
  • the vehicle includes the body, power components, processor, internal bus, network interface, memory, and non-volatile memory, and of course may also include hardware required by other services.
  • the processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, so as to realize the vehicle position prediction method described in any of the above embodiments.
  • the present application also provides a computer program product, including a computer program.
  • the computer program When the computer program is executed by a processor, it can be used to perform one of the methods described in any of the above embodiments. Prediction method of vehicle position.
  • the present application also provides a computer storage medium, where a computer program is stored in the storage medium, and when the computer program is executed by a processor, it can be used to execute the method described in any of the above embodiments.
  • a method for predicting vehicle position is also provided.
  • the device embodiment since it basically corresponds to the method embodiment, for related parts, please refer to the part description of the method embodiment.
  • the device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in One place, or it can be distributed to multiple network elements. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. It can be understood and implemented by those skilled in the art without creative effort.

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Traffic Control Systems (AREA)
  • Navigation (AREA)

Abstract

一种车辆位置的预测方法、装置、车辆及存储介质，获取目标车辆当前所处的路段和车道(110)，以及车道拓扑信息，其中，目标车辆可以是本车周围的可能会影响到本车行为的车辆。而车道拓扑信息用于指示多个连续路段中每一路段的任一车道，与该路段相邻的下一个路段中的车道是否可直接通行(120)。通过目标车辆所处的路段以及车道拓扑信息来预测目标车辆在下一路段所处的车道(130)，能够准确地预测道路中其他车辆的意图。

Description

一种车辆位置的预测方法、装置、车辆及存储介质 技术领域
本申请涉及汽车技术领域，尤其涉及一种车辆位置的预测方法、装置、车辆及存储介质。
背景技术
随着自动驾驶领域的高速发展，自动驾驶技术的智能性、自主性日渐提高，其应用场景也愈发丰富。在自动驾驶的过程中，交通参与者，包括行人与车辆等可移动的物体，他们的行为意图会影响到本车自动驾驶模块作出的行为规划。如何更准确地预测道路中其他车辆的意图是本领域亟需解决的问题。
发明内容
有鉴于此，本申请的目的之一是提供一种车辆位置的预测方法、装置、车辆及存储介质，能准确地预测道路中其他车辆的意图。
为了达到上述技术效果，本发明实施例公开了如下技术方案：
第一方面，提供了一种车辆位置的预测方法，所述方法包括：
获取目标车辆当前所处的路段和车道；
获取车道拓扑信息，所述车道拓扑信息用于指示：多个连续路段中每一路段的任一车道与该路段相邻的下一路段中的车道是否可直接通行；
基于所述目标车辆当前所处的路段和车道和所述车道拓扑信息，预测所述目标车辆在所述下一路段所处的车道。
第二方面，提供了一种车辆位置的预测装置，包括：
处理器；
用于存储处理器可执行程序指令的存储器；
其中，所述处理器调用所述可执行指令时实现上述第一方面所述方法的操作。
第三方面，提供了一种车辆，包括：
车身；
动力组件；
处理器；
用于存储处理器可执行程序指令的存储器，
其中，所述处理器调用所述可执行指令时实现上述第一方面所述方法的操作。
第四方面，提供了一种计算机程序产品，包括计算机程序，所述计算机程序被处理器执行时实现上述第一方面所述方法的步骤。
第五方面，提供了一种机器可读存储介质，所述机器可读存储介质上存储有若干计算机指令，所述计算机指令被执行时执行上述第一方面所述的方法。
本申请提供的一种车辆位置的预测方法、装置、车辆及存储介质，获取目标车辆当前所处的路段和车道，以及车道拓扑信息。由于车道拓扑信息用于指示多个连续 路段中每一路段的任一车道，与该路段相邻的下一个路段中的车道是否可直接通行，因此可以通过目标车辆所处的路段以及车道拓扑信息来预测目标车辆在下一路段所处的车道，从而能够准确地预测道路中其他车辆的意图。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是本申请根据一实施例示出的一种车辆位置的预测方法的流程图。
图2本申请根据一实施例示出的车道拓扑的示意图。
图3本申请根据另一实施例示出的一种车辆位置的预测方法的流程图。
图4本申请根据另一实施例示出的一种车辆位置的预测方法的流程图。
图5是本申请根据另一实施例示出的一种车辆位置的预测方法的流程图。
图6是本申请根据另一实施例示出的一种车辆位置的预测方法的流程图。
图7是本申请根据一实施例示出的一种车辆位置的预测装置的结构示意图。
图8是本申请根据一实施例示出的一种车辆的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
随着自动驾驶领域的高速发展，自动驾驶技术的智能性、自主性日渐提高，其应用场景也愈发丰富。在自动驾驶领域中，车辆通过搭载的传感器来采集车辆所处环境的环境信息。传感器可以包括视觉传感器(如多个单目或双目视觉装置)、激光雷达、毫米波雷达、惯性测量单元(Inertial Measurement Unit，IMU)、全球导航卫星系统(如全球定位系统Global Positioning System，GPS)、陀螺仪、超声传感器、电子罗盘和气压计等传感器中的一种或多种。传感器所采集的环境信息可以包括用于指示物体距离传感器或距离本车的深度信息、交通指示信息(如交通灯信息、指示牌所指示的信息)、车道位置信息、交通参与者的位置信息等。
在自动驾驶的过程中，交通参与者，包括行人与车辆等除了本车以外的其他可移动物体，他们的行为意图会影响到本车自动驾驶模块作出的行为规划。例如若相邻车道上的车辆变换车道至本车所处的车道上，那么本车的行驶速度、行驶方向等行为将会作出相应的规划。
以车辆为例，车辆的意图可以包括多种，如变道意图、超车意图、减速意图等等。意图的表达方式也包括多种，如利用终点作为车辆意图的表达，又例如利用车间空位作为车辆意图的表达等等。对于本车来说，其他车辆的意图往往难以直接得知，需要利用传感器所采集的环境信息来对其他车辆的意图进行预测，并基于预测结果对 本车的行为进行规划。对其他车辆意图的预测结果的准确度会影响本车的行为规划的准确度。然而在相关技术中，尚未有相关技术能较好地对其他车辆的意图作出准确的预测，这是本领域亟需解决的问题。
在对本实施例展开说明之前，先对相关的概念进行以下解释：
移动物体追踪(Movable Object Tracking，MOT)是基于各种预处理后的传感器数据所获得的在时间上连续的移动物体的检测结果，是对于移动物体过去以及当前的状态的客观表达。一般地，MOT由一个轮廓和一个编号组成。同一个移动物体在不同时间对应的坐标是不同的，但是轮廓和编号应该是相同的。轮廓可以利用坐标系和包围盒(Bounding Box)的方法，或者利用坐标系和多面体/多边形的方法来表示。坐标系在不同的技术方案中有不同的种类选取，直观的选择方法是位姿(Pose)坐标。在一些方案中，为了更好地表达移动物体当前和过去的行为，也会选取位置和/或速度作为坐标。一般来说，由于移动物体的加速度是不可观的，因此很少见选取加速度作为坐标。
运动规划(Motion Planning，MP)是指根据环境信息对本车(Ego Vehicle)的未来的运动进行计划，以达到某些度量(Metric)下的最优。Metric可以包括最小能量，最小加速度，最小速度，无碰撞等。运动规划的输出结果可以是轨迹(Trajectory)，也即进行轨迹规划(Trajectory Planning)。轨迹规划的求解空间为状态空间，也即求解本车在时间参数t对应的状态坐标。状态空间可以包括位置，速度，加速度，曲率等参数。当求解空间只考虑位置时，则称为路径规划(Path Planning)。
行为规划(Behavior Planning)是对本车的未来行为进行规划，用于指导本车的运动规划。行为规划过程包括将各种预处理后的传感器数据(Perception)进行处理后输出行为(Behavior)。Behavior的表达方法可以是参考轨迹，也可以是语义上的行为(如前进，后退，左转等)，还可以是这两者的结合。
轨迹预测(Trajectory Prediction)是指利用MOT预测移动物体的未来若干秒的轨迹。一般来说，是指对除本车以外的其他移动物体的轨迹进行预测。轨迹预测的结果可以指导本车的运动规划。
行为预测(Behavior Prediction)是指对移动物体的未来行为进行预测。一般来说，是指对除本车以外的其他移动物体的行为进行预测。为了在本车的每一次决策规划周期(Planning Cycle)中获得尽可能准确的轨迹预测，可以对移动物体的隐含行为进行预测，这样的预测行为称之为行为预测。
为了提高对其他车辆意图预测结果的准确度，本申请提供了一种车辆位置的预测方法，包括如图1所述的步骤：
步骤110：获取目标车辆当前所处的路段和车道；
步骤120：获取车道拓扑信息，所述车道拓扑信息用于指示：多个连续路段中每一路段的任一车道，与该路段相邻的下一路段中的车道，是否可直接通行；
步骤130：基于所述目标车辆当前所处的路段和车道，和所述车道拓扑信息，预测所述目标车辆在所述下一路段所处的车道。
其中，目标车辆可以是本车周围的可能会影响本车行为规划的车辆。例如可以是在本车前/后方的车辆，也可以是相邻车道上的车辆。在一些实施例中，目标车辆可以是与本车距离在预设的距离范围内的车辆。
车道拓扑信息的获取方式可以有多种，作为例子，可以利用车辆搭载的传感器采集的数据来识别出车道信息，并基于车道信息实时建立车道拓扑信息。其中传感器采集的数据可以是图像数据，车道信息可以包括车道位置信息、车道分布信息、车道数量信息等等。作为另一个例子，可以利用电子地图如高精度地图来获取路网数据，并基于路网数据预先建立车道拓扑信息。
本实施例将道路划分为多个连续路段，同一道路中不同路段的路面信息可能并不相同。例如车道数量、车道之间是否可变道、车辆行驶的限制速度等。路段的划分方式可以包括多种，作为例子，路段可以根据车道数量或者根据车道数量的变化进行划分。如图2所示的车道拓扑示意图，图中示出的圆圈表示拓扑图的节点，本实施例中拓扑图的节点用于表示车道，节点之间的路径表示车道之间的连通情况。作为例子，路段a包括3个车道，路段b包括4个车道。作为例子，路段可以根据车道是否可变道进行划分。例如在连续的两个路段中，前一个路段的车道之间不能变道，相同的车道在后一个路段可以变道。作为例子，路段还可以根据长度进行划分，例如在道路中每50米划分为一个路段。路段的划分方式可以包括但不限于以上列举的例子，也可以是上述例子的组合。
路段可以包括一个或以上的车道。如图2所示，路段a包括车道4、1、7；路段b包括车道5、2、8、10；路段c包括车道6、3；路段d包括车道12、15、16、20；路段e包括车道13、17；路段f包括车道14、18、19；路段g包括车道21、22；路段h包括车道9、11。其中，路段a、b、c是三个连续的路段，路段a、b是相邻的两个路段，路段b、c是相邻的两个路段，路段g、h同样是相邻的路段。相邻的路段可以同属于相同的道路，也可以分别属于不同的道路。
车道拓扑信息用于指示多个连续路段中每一路段的任一车道与该路段相邻的下一路段中的车道是否可直接通行。如图2所示的车道拓扑示意图，示出了路段a中的车道4可以与相邻的下一路段b中的车道5直接通行。又例如，路段e中的车道17可以与相邻的下一路段f中的车道18和车道19直接通行。
由于车道拓扑信息指示了相邻路段中车道的可通行性，因此当获取到目标车辆当前所处的路段与车道后，基于车道拓扑信息可以预测出目标车辆在下一路段可以行驶至的车道，并以目标车辆在下一路段可以行驶至的车道作为目标车辆意图的表达方式。由于可以预测出目标车辆在下一路段可能行驶至的车道，因此可以更准确地对目标车辆进行轨迹预测和行为预测。例如，若本车处在如图2所示的路段b的车道5，而目标车辆当前处在路段d的车道20，那么根据车道拓扑信息可以预测目标车辆可以行驶至下一路段h的车道9。也即可以预测出目标车辆的意图为行驶至路段h的车道9。在预测出目标车辆的意图后，可以进一步地对本车的行为进行规划。如此，与相关技术相比，本申请提出的一种车辆位置的预测方法，首次以目标车辆在下一路段可以行驶至的车道作为车辆意图的表达方式，由于使用这样的表达方式能提高对车辆意图预测的准确度，因此基于车辆意图的预测结果对本车作出的行为规划的准确度也会相应提高，从而提高了行车安全性。
目标车辆当前所处的路段和车道的确定方式可以有多种，在一些实施例中，目标车辆当前所处的路段和车道的确定过程可以包括如图3所示的步骤：
步骤310：获取所述目标车辆的当前位置信息；
步骤320：获取车道的位置信息；
步骤330：基于所述目标车辆的当前位置信息以及所述车道的位置信息，确定所述目标车辆当前所处的路段和车道。
其中，目标车辆的当前位置信息可以是指示目标车辆所处地理位置的信息，如目标车辆的经纬度，或者在世界坐标系下的坐标位置。作为例子，本车通过定位模块可以确定自身当前所处的位置，并基于车辆所搭载的传感器获取目标车辆的深度信息，确定目标车辆与本车的距离。基于自身所处的位置以及与目标车辆的距离，可以确定目标车辆的当前位置信息。
车道的位置信息可以包括车道的分布信息，车道所处的路段信息等等。基于目标车辆当前所处的位置信息以及车道的位置信息，可以确定出目标车辆当前所处的路段和车道。
在另一些实施例中，还可以基于本车当前所处的路段和车道，目标车辆的深度信息，以及车道的位置信息，确定目标车辆当前所处的路段和车道。利用目标车辆的深度信息，可以确定目标车辆与本车的距离，如此根据本车当前所处的路段和车道，可以推算出目标车辆所处的路段和车道。
目标车辆当前所处的路段和车道的确定方式包括但不限于上述两种方式，本领域技术人员可以根据实际需要采取其他方式来确定目标车辆当前所处的路段和车道，本申请在此不做限制。
车道拓扑信息可以指示路段的任一车道与该路段相邻的下一路段的车道是否可直接通行。在一些场景中，若目标车辆当前所处的车道，与相邻的下一路段中两个及以上的车道可以直接通行，那么在预测目标车辆在下一路段所处的车道的过程中，可以包括如图4所示的步骤：
步骤410：基于所述目标车辆当前所处的路段和车道，和所述车道拓扑信息，预测所述目标车辆行驶至所述下一路段的任一车道的预测概率；
步骤420：基于所述预测概率预测所述目标车辆在所述下一路段所处的车道。
如图2所示的车道拓扑示意图中，若目标车辆当前处于路段e中的车道17，根据车道拓扑信息可知车道17可以与相邻的下一路段f中的车道18和车道19直接通行。如此，在预测目标车辆在下一路段所处的车道时，可以预测目标车辆行驶至下一路段任一车道的预测概率，然后基于任一车道的预测概率预测目标车辆在下一路段所处的车道。作为例子，可以将预测概率最大值对应的车道确定为目标车辆在下一路段所处的车道。例如若预测目标车辆行驶至车道18的预测概率为80％，行驶至车道19的预测概率为20％，则预测目标车辆在下一路段f所处的车道为车道18。
在一些场景中，路段可以包括多个车道，车辆在相邻车道可以相互变更车道或者车辆可以从某一车道上行驶到另一相邻的车道上。如此，在预测目标车辆行驶至下一路段的任一车道的预测概率时，不仅需要预测当前车道可以直接通行的下一路段的车道的预测概率，还需要预测与当前车道相邻的车道可以直接通行的下一路段的车道的预测概率。如此，上述预测目标车辆行驶至下一路段的任一车道的预测概率，可以包括如图5所示的步骤：
步骤510：基于当前车道，所述当前车道的相邻车道，以及所述车道拓扑信息，预测所述目标车辆经过的预测车道序列组合；
步骤520：预测所述目标车辆经过所述预测车道序列组合的预测概率。
其中，预测车道序列组合至少由当前路段上的车道以及当前路段的相邻路段的车道按照先后排序组成。作为例子，预测车道序列组合可以包括当前车道与相邻路段中可直接通行的车道按照先后排序组成的车道序列组合。以图2为例，若目标车辆A当前处于车道1，根据车道拓扑信息，可知车道2可以与车道1直接通行，车道3可以与车道2直接通行，则预测车道序列组合可以包括车道1、车道2以及车道3按照先后排序组成的车道序列组合，表示为车道序列1-2-3。
作为例子，预测车道序列组合也可以包括相邻车道与相邻路段中可直接通行的车道按照先后排序组成的车道序列组合。其中，相邻车道可以包括与目标车辆当前所处车道相邻的左右两个车道中的一个或两个。如在上述例子中，目标车辆A当前所处的车道1的相邻车道包括车道4和车道7。根据车道拓扑信息，可知车道4、车道5以及车道6可以依次直接通行。如此，预测车道序列组合可以包括车道4、车道5以及车道6按照先后排序组成的车道序列组合，表示为车道序列4-5-6。同理，预测车道序列组合还可以包括车道7、车道8以及车道9按照先后排序组成的车道序列组合，表示为车道序列7-8-9。以及预测车道序列组合还可以包括车道7、车道10以及车道11按照先后排序组成的车道序列组合，表示为车道序列7-10-11。
如此，在预测处于车道1的目标车辆A行驶至下一路段任一车道的预测概率时，可以预测车道序列1-2-3的预测概率，以及可以预测车道序列4-5-6、车道序列7-8-9、车道序列7-10-11中的一个或多个车道序列的预测概率。从而覆盖了目标车辆可能行驶至的车道的各种可能性，提高目标车辆意图预测的准确度。
在上述任一实施例中，目标车辆行驶至下一路段任一车道的预测概率，和/或目标车辆经过预测车道序列组合的预测概率，可以是根据经验值人工指定的一个概率分布先验，也可以是通过车道序列组合的历史概率获取的。某一时刻下目标车辆的车道序列组合及其预测概率的表示形式可以包括多种，如在上述例子中，可以通过人工指定或历史概率为各个车道序列赋予一个初始的概率值，表示为：1-2-3(P1)/4-5-6(P2)/7-8-9(P3)/7-10-11(P4)。其中，P1-P4为各车道序列的预测概率值。
随着目标车辆的行驶，当目标车辆到达下一个路段或者变更车道时，目标车辆行驶至下一路段任一车道的预测概率，以及目标车辆经过预测车道序列组合的预测概率也会随之改变。在更新预测概率时，可以在时序上利用历史概率，以提高预测概率的准确性。如此，在一些实施例中，预测概率还可以基于历史概率预测，那么上述任一实施例还可以包括步骤：获取预设时间间隔之前时刻确定的目标车辆经过历史车道序列组合的历史概率。其中，预设时间间隔可以根据实际需要进行设定，如1秒或3秒。历史车道序列组合至少由目标车辆预设时间间隔之前时刻经过的历史路段上的车道以及历史路段的相邻路段的车道按照先后排序组成。如在上述例子中，若目标车辆A在第一时刻处于车道1，那么在第一时刻的预测车道序列组合包括：1-2-3(P1)/4-5-6(P2)/7-8-9(P3)/7-10-11(P4)。当目标车辆A在第二时刻行驶到路段b的车道2时，那么第一时刻为第二时刻的预设时间间隔之前的时刻，第一时刻的预测车道序列组合为第二时刻的历史车道序列组合。基于第一时刻的预测车道序列组合的预测概率，可以预测第二时刻的预测车道序列组合的预测概率。其中，第二时刻的预测车道序列可以包括：2-3(P5)/5-6(P6)/8-9(P7)。
在一些实施例中，若目标车辆当前所处的路段与预设时间间隔之前时刻确定的目标车辆所处的历史路段相同，将历史概率确定为目标车辆驶向预测车道序列组合的预测概率。如在上述例子中，目标车辆A在第一时刻处于车道1，在经过预设时间间隔后目标车辆A依然处在车道1，那么该时刻的预测车道序列组合及其的预测概率依然为第一时刻的预测车道序列组合及其预测概率。
在一些实施例中，若目标车辆当前所处的路段为预设时间间隔之前时刻确定的目标车辆所处的历史路段的延伸，根据车道拓扑信息对历史车道序列组合的历史概率进行分配得到预测车道序列组合的预测概率。如在上述例子中，目标车辆A在第二时刻处于车道2，则第二时刻的预测车道序列包括：2-3(P5)/5-6(P6)/8-9(P7)。其中车道序列2-3为第一时刻的车道序列1-2-3的延伸，则确定第二时刻车道序列2-3的预测概率P5继承第一时刻车道序列1-2-3的预测概率P1。同理，车道序列5-6的预测概率P6继承车道序列4-5-6的预测概率P2，车道序列8-9的预测概率P7继承车道序列7-8-9的预测概率P3。
在一些实施例中，若当前路段与历史路段之间存在非一一对应的车道通行关系，将历史车道序列组合的历史概率按照车道通行关系进行分配得到预测车道序列组合的预测概率。其中，非一一对应的车道通行关系包括历史路段的一个车道与当前路段两个以上的车道可直接通行。如图2中车道7可与车道8以及车道10直接通行。此外，还包括历史路段两个以上的车道与当前路段的一个车道可直接通行。如图2中车道12和车道15均可与车道13直接通行。
若上述车道通行关系为：历史路段的一个车道与当前路段两个以上的车道可直接通行，可以按照最小熵原则对历史概率进行分配得到预测车道序列组合的预测概率。
若上述车道通行关系为：历史路段两个以上的车道与当前路段的一个车道可直接通行，当前路段的一个车道对应的车道序列继承该历史路段两个以上车道对应的车道序列的历史概率。
在本实施例中，考虑到车道拓扑在时序上可延续，因此在对当前的预测车道序列组合进行预测时，结合了历史车道序列组合的历史概率进行预测。由于沿用了历史概率，目标车辆的行为预测的准确度得以大大提高。
在一些实施例中，上述任一实施例的预测概率还可以基于目标车辆的运动状态信息进行预测。目标车辆的运动状态信息可以基于MOT获取。在一些实施例中，目标车辆的预测车道序列组合的预测概率可以基于上述的历史概率以及运动状态信息获取。作为例子，可以基于运动状态信息以及预设更新策略对历史概率进行迭代更新，直至更新过程中所得概率的变化小于预设变化阈值。所得概率称为后验概率，将后验概率确定为预测车道序列组合的预测概率。
在一些实施例中，上述预设更新策略可以包括贝叶斯更新。
在一些实施例中，上述任一实施例还可以包括步骤：基于所预测的目标车辆在下一路段所处的车道，对车辆进行行为规划。
如此，在确定预测车道序列组合的预测概率时，结合了历史车道序列组合的历史概率以及目标车辆当前的运动状态信息进行预测。运动状态信息在一定程度上可以对基于历史概率预测出的预测概率进行修正，进一步地提高了预测概率的准确度。在对目标车辆的预测概率准确度得以提高的情况下，本车相应的行为规划的准确度也能 大大提高。
作为例子，如上所述，目标车辆A在第一时刻处在车道1，第一时刻的预测车道序列组合包括：1-2-3(P1)/4-5-6(P2)/7-8-9(P3)/7-10-11(P4)。基于经验值，目标车辆A保持行驶在当前车道1的概率为0.6，左/右变道的概率均为0.2。可知，P1＝0.6，P2＝0.2，P3+P4＝0.2。按照最小熵原则对右变道的概率进行分配，得到P3＝P4＝0.1。最终可知第一时刻的预测车道序列组合包括：1-2-3(0.6)/4-5-6(0.2)/7-8-9(0.1)/7-10-11(0.1)。
若目标车辆A在第二时刻行驶到路段b的车道2，第二时刻的预测车道序列组合包括：2-3(P5)/5-6(P6)/8-9(P7)。如上所述，车道序列2-3的预测概率P5继承第一时刻车道序列1-2-3的预测概率P1，车道序列5-6的预测概率P6继承车道序列4-5-6的预测概率P2。车道序列8-9的预测概率P7继承车道序列7-8-9的预测概率P3。将各预测概率P5-P3进行归一化处理后，最终可知第二时刻的预测车道序列组合包括：2-3(0.67)/5-6(0.22)/8-9(0.11)。
若目标车辆A在第三时刻更换车道至车道8，第三时刻的预测车道序列组合包括：2-3(P5)/8-9(P7)/10-11(P8)。由于车道序列10-11在第二时刻未存在历史路段对应的历史车道序列组合，因此可以根据经验值得知右变道的概率为0.2，即人工指定车道序列10-11的预测概率P8为0.2。而第三时刻的车道序列2-3以及车道序列8-9分别与第二时刻的车道序列2-3以及车道序列8-9相同，因此第三时刻的车道序列2-3的预测概率以及车道序列8-9的预测概率直接继承第二时刻的车道序列2-3的预测概率以及车道序列8-9的预测概率。经过归一化处理后，最终第三时刻的预测车道序列组合包括：2-3(0.68)/8-9(0.11)/10-11(0.21)。
作为另一个例子，如图2所示，若目标车辆B在第一时刻处在车道15，则第一时刻的预测车道序列组合包括：15-13-14(P9)/12-13-14(P10)/16-17-18(P11)/16-17-19(P12)。基于经验值，目标车辆B保持行驶在当前车道15的概率为0.6，左/右变道的概率均为0.2。可知，P9＝0.6，P10＝0.2，P11+P12＝0.2。按照最小熵原则对右变道的概率进行分配，得到P11＝P12＝0.1。最终可知第一时刻的预测车道序列组合包括：15-13-14(0.6)/12-13-14(0.2)/16-17-18(0.1)/16-17-19(0.1)。
若目标车辆B在第二时刻行驶至车道13，则第二时刻的预测车道序列组合包括13-14(P13)/17-18(P14)/17-19(P15)。由于第二时刻的历史路段d的车道12与车道15与第三时刻的车道13可直接通行，因此车道序列13-14的预测概率P13继承车道序列15-13-14的预测概率P9以及车道序列12-13-14的预测概率P10。而第三时刻的车道序列17-18为第二时刻的车道序列16-17-18的延伸；第三时刻的车道序列17-19为第二时刻的车道序列16-17-19的延伸。因此车道序列17-18的预测概率P14继承车道序列16-17-18的预测概率P11；车道序列17-19的预测概率P15继承车道序列16-17-19的的预测概率P12。将各预测概率进行归一化处理后，最终可知第二时刻的预测车道序列组合包括13-14(0.8)/17-18(0.1)/17-19(0.1)。
本申请提供的一种车辆位置的预测方法，获取目标车辆当前所处的路段和车道，以及车道拓扑信息，由于车道拓扑信息用于指示多个连续路段中每一路段的任一车道，与该路段相邻的下一个路段中的车道是否可直接通行，即车道拓扑信息指示了相邻路段中车道的可通行性，因此基于车道拓扑信息可以预测出目标车辆在下一路段 可以通行的车道。
与相关技术相比，本申请首次以目标车辆在下一路段可以行驶至的车道作为车辆意图的表达方式。同时，由于车道拓扑在时序上可延续，因此在对当前的预测车道序列组合进行预测时，可以结合历史车道序列组合的历史概率进行预测。由于沿用了历史概率，目标车辆的行为预测的准确度得以大大提高。
同时在确定预测车道序列组合的预测概率时，还结合了目标车辆当前的运动状态信息进行预测，运动状态信息在一定程度上可以对基于历史概率预测出的预测概率进行修正，进一步地提高了预测概率的准确度。在对目标车辆的预测概率准确度得以提高的情况下，本车相应的行为规划的准确度也能大大提高。
此外，本申请还提供了一种车辆位置的预测方法，包括如图6所示的步骤：
步骤610：获取目标车辆的当前位置信息以及车道的位置信息；
步骤620：基于所述目标车辆的当前位置信息以及所述车道的位置信息，确定所述目标车辆当前所处的路段和车道；
步骤630：获取车道拓扑信息，所述车道拓扑信息用于指示：多个连续路段中每一路段的任一车道，与该路段相邻的下一路段中的车道，是否可直接通行；
步骤640：基于当前车道，所述当前车道的相邻车道，以及所述车道拓扑信息，预测所述目标车辆经过的预测车道序列组合；
其中，所述预测车道序列组合至少由当前路段上的车道以及当前路段的相邻路段的车道按照先后排序组成。
步骤650：获取预设时间间隔之前时刻确定的所述目标车辆经过历史车道序列组合的历史概率；
其中，所述历史车道序列组合至少由所述目标车辆预设时间间隔之前时刻经过的历史路段上的车道以及所述历史路段的相邻路段的车道按照先后排序组成。
步骤660：获取所述目标车辆的运动状态信息；
步骤670：基于所述运动状态信息以及贝叶斯更新对所述历史概率进行迭代更新，直至所得概率的变化小于预设变化阈值，所得概率为所述预测车道序列组合的预测概率；
步骤680：基于所述预测概率预测所述目标车辆在所述下一路段所处的车道；
步骤690：基于预测的所述目标车辆在所述下一路段所处的车道，对车辆进行行为规划。
具体实现方式参见上文实施例，本申请在此不再赘述。
本申请提供的一种车辆位置的预测方法，获取目标车辆当前所处的路段和车道，以及车道拓扑信息，其中，目标车辆可以是本车周围的可能会影响到本车行为的车辆。而车道拓扑信息用于指示多个连续路段中每一路段的任一车道，与该路段相邻的下一个路段中的车道是否可直接通行。由于车道拓扑信息指示了相邻路段中车道的可通行性，因此基于车道拓扑信息可以预测出目标车辆在下一路段可以行驶至的车道。
与相关技术相比，本申请首次以目标车辆在下一路段可以行驶至的车道作为车辆意图的表达方式。同时，由于车道拓扑在时序上可延续，因此在对当前的预测车道序列组合进行预测时，可以结合历史车道序列组合的历史概率进行预测。由于沿用了 历史概率，目标车辆的行为预测的准确度得以大大提高。
同时在确定预测车道序列组合的预测概率时，还结合了目标车辆当前的运动状态信息进行预测，运动状态信息在一定程度上可以对基于历史概率预测出的预测概率进行修正，进一步地提高了预测概率的准确度。在对目标车辆的预测概率准确度得以提高的情况下，本车相应的行为规划的准确度也能大大提高。
基于上述任意实施例所述的一种车辆位置的预测方法，本申请还提供了如图7所示的一种车辆位置的预测装置的结构示意图。如图7，在硬件层面，该预测装置包括处理器、内部总线、网络接口、内存以及非易失性存储器，当然还可能包括其他业务所需要的硬件。处理器从非易失性存储器中读取对应的计算机程序到内存中然后运行，以实现上述任意实施例所述的一种车辆位置的预测方法。
基于上述任意实施例所述的一种车辆位置的预测方法，本申请还提供了如图8所示的一种车辆的结构示意图。如图8，在硬件层面，该车辆包括车身、动力组件、处理器、内部总线、网络接口、内存以及非易失性存储器，当然还可能包括其他业务所需要的硬件。其中，处理器从非易失性存储器中读取对应的计算机程序到内存中然后运行，以实现上述任意实施例所述的一种车辆位置的预测方法。
基于上述任意实施例所述的一种车辆位置的预测方法，本申请还提供了一种计算机程序产品，包括计算机程序，计算机程序被处理器执行时可用于执行上述任意实施例所述的一种车辆位置的预测方法。
基于上述任意实施例所述的一种车辆位置的预测方法，本申请还提供了一种计算机存储介质，存储介质存储有计算机程序，计算机程序被处理器执行时可用于执行上述任意实施例所述的一种车辆位置的预测方法。
对于装置实施例而言，由于其基本对应于方法实施例，所以相关之处参见方法实施例的部分说明即可。以上所描述的装置实施例仅仅是示意性的，其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性劳动的情况下，即可以理解并实施。
需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
以上对本申请实施例所提供的方法和装置进行了详细介绍，本文中应用了具体个例对本申请的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本申请的方法及其核心思想；同时，对于本领域的一般技术人员，依据本申请的思想，在具体实施方式及应用范围上均会有改变之处，综上所述，本说明书内容不应理解为 对本申请的限制。

Claims (16)

  1. 一种车辆位置的预测方法，其特征在于，所述方法包括：
    获取目标车辆当前所处的路段和车道；
    获取车道拓扑信息，所述车道拓扑信息用于指示：多个连续路段中每一路段的任一车道与该路段相邻的下一路段中的车道是否可直接通行；
    基于所述目标车辆当前所处的路段和车道和所述车道拓扑信息，预测所述目标车辆在所述下一路段所处的车道。
  2. 根据权利要求1所述的方法，其特征在于，所述获取目标车辆当前所述的路段和车道，包括：
    获取所述目标车辆的当前位置信息；
    获取车道的位置信息；
    基于所述目标车辆的当前位置信息以及所述车道的位置信息，确定所述目标车辆当前所处的路段和车道。
  3. 根据权利要求1所述的方法，其特征在于，所述基于所述目标车辆当前所处的路段和车道，和所述车道拓扑信息，预测所述目标车辆在所述下一路段所处的车道，包括：
    基于所述目标车辆当前所处的路段和车道，和所述车道拓扑信息，预测所述目标车辆行驶至所述下一路段的任一车道的预测概率；
    基于所述预测概率预测所述目标车辆在所述下一路段所处的车道。
  4. 根据权利要求3所述的方法，其特征在于，所述基于所述目标车辆当前所处的路段和车道，和所述车道拓扑信息，预测所述目标车辆行驶至所述下一路段的任一车道的预测概率，包括：
    基于当前车道，所述当前车道的相邻车道，以及所述车道拓扑信息，预测所述目标车辆经过的预测车道序列组合，所述预测车道序列组合至少由当前路段上的车道以及所述当前路段的相邻路段的车道按照先后排序组成；
    预测所述目标车辆经过所述预测车道序列组合的预测概率。
  5. 根据权利要求4所述的方法，其特征在于，所述方法还包括：
    获取预设时间间隔之前时刻确定的所述目标车辆经过历史车道序列组合的历史概率，所述历史车道序列组合至少由所述目标车辆预设时间间隔之前时刻经过的历史路段上的车道以及所述历史路段的相邻路段的车道按照先后排序组成；
    所述预测概率还基于所述历史概率预测。
  6. 根据权利要求5所述的方法，其特征在于，所述预测概率的预测过程包括：
    如果所述目标车辆当前所处的路段与预设时间间隔之前时刻确定的所述目标车辆所处的历史路段相同，将所述历史概率确定为所述目标车辆驶向所述预测车道序列组合的预测概率。
  7. 根据权利要求5所述的方法，其特征在于，所述预测概率的预测过程包括：
    如果所述目标车辆当前所处的路段为预设时间间隔之前时刻确定的所述目标车辆所处的历史路段的延伸，根据所述车道拓扑信息对所述历史车道序列组合的历史概率进行分配得到所述预测车道序列组合的预测概率。
  8. 根据权利要求5所述的方法，其特征在于，所述预测概率的预测过程包括：
    如果当前路段与历史路段之间存在非一一对应的车道通行关系，将所述历史车道序列组合的历史概率按照所述车道通行关系进行分配得到所述预测车道序列组合的预测概率。
  9. 根据权利要求5-8任一所述的方法，其特征在于，所述方法还包括：
    获取所述目标车辆的运动状态信息；
    所述预测概率还基于所述运动状态信息预测。
  10. 根据权利要求9所述的方法，其特征在于，所述预测概率的预测过程包括：
    基于所述运动状态信息以及预设更新策略对所述历史概率进行迭代更新，直至更新过程中所得概率的变化小于预设变化阈值，将所得概率确定为预测概率。
  11. 根据权利要求10所述的方法，其特征在于，所述预设更新策略为贝叶斯更新。
  12. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    基于所述预测的所述目标车辆在所述下一路段所处的车道，对车辆进行行为规划。
  13. 一种车辆位置的预测装置，其特征在于，包括：
    处理器；
    用于存储处理器可执行程序指令的存储器；
    其中，所述处理器调用所述可执行指令时实现权利要求1-12任一所述方法的操作。
  14. 一种车辆，其特征在于，包括：
    车身；
    动力组件；
    处理器；
    用于存储处理器可执行程序指令的存储器，
    其中，所述处理器调用所述可执行指令时实现如权利要求1-12任一所述方法的操作。
  15. 一种计算机程序产品，包括计算机程序，其特征在于，所述计算机程序被处理器执行时实现如权利要求1-12任一所述方法的步骤。
  16. 一种机器可读存储介质，其特征在于，所述机器可读存储介质上存储有若干计算机指令，所述计算机指令被执行时执行权利要求1-12任一所述的方法。
PCT/CN2021/143925 2021-12-31 2021-12-31 一种车辆位置的预测方法、装置、车辆及存储介质 WO2023123456A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2021/143925 WO2023123456A1 (zh) 2021-12-31 2021-12-31 一种车辆位置的预测方法、装置、车辆及存储介质
CN202180102020.8A CN117897749A (zh) 2021-12-31 2021-12-31 一种车辆位置的预测方法、装置、车辆及存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/143925 WO2023123456A1 (zh) 2021-12-31 2021-12-31 一种车辆位置的预测方法、装置、车辆及存储介质

Publications (1)

Publication Number Publication Date
WO2023123456A1 true WO2023123456A1 (zh) 2023-07-06

Family

ID=86997162

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/143925 WO2023123456A1 (zh) 2021-12-31 2021-12-31 一种车辆位置的预测方法、装置、车辆及存储介质

Country Status (2)

Country Link
CN (1) CN117897749A (zh)
WO (1) WO2023123456A1 (zh)

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150266477A1 (en) * 2014-03-24 2015-09-24 Honda Research Institute Europe Gmbh Method and system for predicting movement behavior of a target traffic object
CN105593643A (zh) * 2013-10-09 2016-05-18 丰田自动车株式会社 车辆用车道引导系统和车辆用车道引导方法
CN107919027A (zh) * 2017-10-24 2018-04-17 北京汽车集团有限公司 预测车辆变道的方法、装置和系统
CN109186617A (zh) * 2018-08-13 2019-01-11 武汉中海庭数据技术有限公司 一种基于视觉众包数据自动生成车道级拓扑关系的方法、系统及存储器
CN110400490A (zh) * 2019-08-08 2019-11-01 腾讯科技（深圳）有限公司 轨迹预测方法和装置
CN110502012A (zh) * 2019-08-20 2019-11-26 武汉中海庭数据技术有限公司 一种他车轨迹预测方法、装置及存储介质
US20200172093A1 (en) * 2018-11-29 2020-06-04 291, Daehak-ro Lane-based probabilistic motion prediction of surrounding vehicles and predictive longitudinal control method and apparatus
CN111523706A (zh) * 2020-04-06 2020-08-11 武汉理工大学 一种基于深度学习组合模型的断面车道级短时交通流预测方法
CN112833903A (zh) * 2020-12-31 2021-05-25 广州文远知行科技有限公司 一种轨迹预测方法、装置、设备和计算机可读存储介质
CN112884014A (zh) * 2021-01-26 2021-06-01 南京邮电大学 一种基于路段拓扑结构分类的交通速度短时预测方法

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105593643A (zh) * 2013-10-09 2016-05-18 丰田自动车株式会社 车辆用车道引导系统和车辆用车道引导方法
US20150266477A1 (en) * 2014-03-24 2015-09-24 Honda Research Institute Europe Gmbh Method and system for predicting movement behavior of a target traffic object
CN107919027A (zh) * 2017-10-24 2018-04-17 北京汽车集团有限公司 预测车辆变道的方法、装置和系统
CN109186617A (zh) * 2018-08-13 2019-01-11 武汉中海庭数据技术有限公司 一种基于视觉众包数据自动生成车道级拓扑关系的方法、系统及存储器
US20200172093A1 (en) * 2018-11-29 2020-06-04 291, Daehak-ro Lane-based probabilistic motion prediction of surrounding vehicles and predictive longitudinal control method and apparatus
CN110400490A (zh) * 2019-08-08 2019-11-01 腾讯科技（深圳）有限公司 轨迹预测方法和装置
CN110502012A (zh) * 2019-08-20 2019-11-26 武汉中海庭数据技术有限公司 一种他车轨迹预测方法、装置及存储介质
CN111523706A (zh) * 2020-04-06 2020-08-11 武汉理工大学 一种基于深度学习组合模型的断面车道级短时交通流预测方法
CN112833903A (zh) * 2020-12-31 2021-05-25 广州文远知行科技有限公司 一种轨迹预测方法、装置、设备和计算机可读存储介质
CN112884014A (zh) * 2021-01-26 2021-06-01 南京邮电大学 一种基于路段拓扑结构分类的交通速度短时预测方法

Also Published As

Publication number Publication date
CN117897749A (zh) 2024-04-16

Similar Documents

Publication Publication Date Title
EP3699048A1 (en) Travelling track prediction method and device for vehicle
US20240144010A1 (en) Object Detection and Property Determination for Autonomous Vehicles
US11545033B2 (en) Evaluation framework for predicted trajectories in autonomous driving vehicle traffic prediction
CN110857085A (zh) 车辆路径规划
US20210173402A1 (en) Systems and methods for determining vehicle trajectories directly from data indicative of human-driving behavior
US11055540B2 (en) Method for determining anchor boxes for training neural network object detection models for autonomous driving
US11351996B2 (en) Trajectory prediction of surrounding vehicles using predefined routes
US20210389133A1 (en) Systems and methods for deriving path-prior data using collected trajectories
US20220101155A1 (en) Trajectory Generation Using Road Network Model
CN110096054A (zh) 用于使用多个线程生成用于自动驾驶车辆的参考线的方法和系统
RU2757234C2 (ru) Способ и система для вычисления данных для управления работой беспилотного автомобиля
US11467584B2 (en) Multi-layer grid based open space planner
US20220234575A1 (en) Vehicle operation using maneuver generation
CN114061581A (zh) 通过相互重要性对自动驾驶车辆附近的智能体排名
US11577732B2 (en) Methods and systems for tracking a mover's lane over time
GB2599191A (en) Monocular 3D object detection from image semantics network
JP2023523350A (ja) 乗り物に基づくデータ処理方法、データ処理装置、コンピュータ機器、及びコンピュータプログラム
US20230333237A1 (en) Merge-split techniques for sensor data filtering
JP6507841B2 (ja) 先行車両推定装置及びプログラム
WO2023097874A1 (zh) 用于行驶轨迹规划的方法和装置
US20210181738A1 (en) Method for determining ride stability of an autonomous driving system controlling an autonomous driving vehicle
WO2023123456A1 (zh) 一种车辆位置的预测方法、装置、车辆及存储介质
CN113753038B (zh) 一种轨迹预测方法、装置、电子设备和存储介质
EP4134623A1 (en) Drive device, vehicle, and method for automated driving and/or assisted driving
US11946749B2 (en) Driving data guided spatial planning

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21969801

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 202180102020.8

Country of ref document: CN