IMPROVED METHOD FOR DETERMINING THE EDGE STRUCTURES OF A ROAD FOR THE LOCALIZATION OF MOTOR VEHICLES
Patent Information
- Application Number
- DE502016016762
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2015-12-17
- Filing Date
- 2016-11-14
- Publication Date
- 2024-11-14
- Estimated Expiration
- 2036-11-14
AI Technical Summary
Current methods for vehicle localization on roads, especially in highly automated or autonomous vehicles, face challenges with the accuracy of satellite-based systems and require complex and costly high-precision road models that need frequent updates due to environmental changes.
A method that uses a network of vehicles with standard GNSS systems and sensors to collaboratively determine the position of peripheral road features, such as guardrails and signs, by calculating probability distributions of object positions and fusing data from multiple vehicles to achieve higher accuracy without the need for specialized equipment.
This approach allows for precise and cost-effective determination of road features, enabling accurate vehicle positioning and quick updates without expensive sensors, improving accuracy and stability against environmental disturbances.
Description
[0001] The invention relates to an improved method for locating vehicles on a roadway by evaluating the edge buildings and a method for determining the edge buildings.
[0002] Highly automated or fully autonomous vehicles must accurately determine their position on the road. The accuracy of conventional satellite-based systems (GNSS-based systems) is insufficient to determine a vehicle's position on the road with sufficient precision.
[0003] Therefore, the prior art proposes that vehicles determine their position on the road relative to the roadside buildings. This requires systems for creating highly accurate road models. Such road models contain significantly more information than the maps currently used in mass-produced navigation systems. The road models include information about the number of lanes, the type of lane markings, the type and location of roadside buildings, the type and location of signage, etc. In the prior art, such road models are typically acquired by sensor vehicles equipped with precise and complex measuring instruments. These sensor vehicles have very accurate localization systems, such as GNSS-based inertial platforms, as well as high-resolution sensors for capturing the surroundings, such as a camera and a laser scanner.Such systems are revealed, for example, in Vehicle Self-localization with High-Precision Digital Maps, Andreas Schindler, IEEE Intelligent Vehicles Symposium, 2013 and Generating High Precision Maps for Advanced Guidance Support, Ulf Noyer et al., IEEE Intelligent Vehicles Symposium, 2008, Collective Learning of High-Precision Road Models, Sebastian Rauch, 6th Conference on Driver Assistance, Munich, 2013.
[0004] To locate a vehicle with lane-level precision while driving, especially on a motorway or other multi-lane highway, a highly accurate digital map and high-precision sensors are used. Such methods are currently in the experimental stage.
[0005] The dissertation Vehicle Self-localization Using High-Precision Digital Maps, Andreas Schindler, FORWISS 2013, reveals a method that also requires such a highly accurate road model, using simple GNSS data and camera data for localizing the vehicle while driving.
[0006] State-of-the-art methods and systems require a highly precise road model for lane-accurate vehicle localization. This model is generated using sensor vehicles equipped with complex sensors. Such a road model has the disadvantage that measurement data must be reacquired whenever changes are made to the road's surrounding structures. Furthermore, state-of-the-art methods and systems have the disadvantage that, depending on the number of lanes, surrounding traffic, weather conditions, etc., multiple passes over a road segment may be necessary to capture the surrounding structures. Another disadvantage of state-of-the-art systems and methods is that the sensor vehicle must retraceive a road segment to verify whether the highly precise road model is still up-to-date or has become outdated, for example, due to structural changes.
[0007] CN 101206644 B discloses the central collection of data obtained from vehicles.
[0008] German patent DE 10 2013 208 521 A1 discloses a method for the collective learning and creation of a digital road model. Trajectory and perception data are collected from a plurality of vehicles. Associations between the trajectories are created by forming feature grids and probability field feature grids for corresponding trajectory points to be associated and correlating these to generate association hypotheses.
[0009] The invention aims to provide an improved method for generating road models and locating a vehicle.
[0010] The object of the invention is achieved by a method according to one of the independent claims. The dependent claims specify preferred embodiments.
[0011] A method according to the invention for estimating the position of the buildings along the edge of a road comprises the step of driving a plurality of first vehicles along a road segment, each moving along a trajectory, wherein each of the plurality of first vehicles performs the steps described below. Each first vehicle determines a first value of a satellite-based position of the respective first vehicle using a satellite-based positioning system with position determination means of the respective first vehicle. The first value has a first accuracy. In other words, each first vehicle determines its position using a satellite-based system, for example a GNSS system, as it moves along the trajectory. The accuracy of the first value is limited by the accuracy of the satellite-based system.
[0012] Each vehicle determines the position of the edge building by means of a measurement with a sensor on the vehicle and based on the satellite-based position of the first vehicle. The determined position of the edge building is sent to a central facility, for example, a backend.
[0013] The procedure also includes the steps of receiving multiple determined positions of the edge buildings, each determined by one of multiple vehicles, at the central facility, and determining the probable position of the edge buildings based on these multiple determined positions. By recording the positions of the edge buildings from multiple vehicles using their onboard sensors and linking them with satellite-based vehicle positions, each vehicle can determine values for the position of the edge buildings. At the central facility, the positions determined by the multiple vehicles are processed to determine the probable position of the edge buildings, for example, with higher accuracy. The probable position of the edge buildings is then transmitted to at least one vehicle.
[0014] Each first vehicle determines initial object position values Zk = {(xn, yn)|n = 1,..., N} as the determined position of a boundary structure at time k of a plurality N of static objects Zk outside the roadway relative to the respective first vehicle, using the sensor of that first vehicle. The initial object position values are determined at a plurality of times k, k = 1,..., K. The initial object position values are also determined based on the satellite-based position of the first vehicle, but this is not necessarily required. For example, in one embodiment, a coordinate transformation can be used to determine the initial object position values in a world coordinate system.In another embodiment, the first values can be determined in the sensor coordinate system, the vehicle coordinate system, or any other quasi-stationary or moving coordinate system. The first values of the object positions Zk exhibit a second accuracy. This second accuracy is determined, firstly, by the first accuracy of the satellite-based position and, secondly, by the accuracy of the sensor for detecting the perimeter buildings. The determined position of the perimeter buildings is transmitted to a central unit. This central unit can be a so-called backend.
[0015] The procedure determines a first probability distribution P(S|ZK) that indicates whether a static object of a perimeter building is located at the second values of object positions (xi, yi) of a detected plane, based on the plurality K determinations of the first values of object positions Zk at times k for k = 1,..., K, where P(S|ZK) is the conditional probability of a perimeter building S = (xi, yi) based on the measurements Z, where Z comprises the (x, y) point cloud, and ZK comprises all previous determinations at times k of object positions Zk and the current determination at time K of object positions ZK by the first vehicle (k = 1,..., K). In other words, an object of the perimeter building is detected by the sensor at a plurality of times k and can subsequently be transformed into a world coordinate system or another coordinate system. These values are referred to as the first values of the object positions.In the step of determining an initial probability distribution, the procedure evaluates all determined initial values of object positions Zk to determine the probable location S of an object in the perimeter development. This reduces inaccuracies due to the sensor. The initial probability distribution is determined by the first vehicle and transmitted to a central unit as the determined position of the perimeter development.
[0016] However, it is also possible outside of the method according to the invention that the first probability distribution is determined by a central unit outside the first vehicle.
[0017] The procedure also includes receiving at the central facility a plurality of first probability distributions for the probable position of the edge buildings, each determined by one of a plurality of vehicles, as the determined position of the edge buildings, and determining the probable position of the edge buildings based on the plurality of first probability distributions. Since each vehicle already determines the probable position of the edge buildings, higher accuracy and greater stability against disturbances can be achieved.
[0018] The procedure determines third values of the object positions mfusion as the probable position of the perimeter building based on a plurality of first probability distributions, each determined by a plurality of vehicles. The third values of the object positions mfusion of the perimeter building can have a higher accuracy than the first values of the object positions (xn, yn) and a higher accuracy than the second values of the object positions (xi, yi). This step can be performed by the central unit. The first vehicles can send the first values of the object positions or, preferably, the first probability distribution to the central unit. From the plurality of first probability distributions, each determined by different vehicles, the third values of the object positions mfusion of the perimeter building position are generated.
[0019] The present invention has the advantage that the position of roadside structures can be determined with very high accuracy without requiring specialized vehicles equipped with expensive sensors. The inventors of the present invention have proposed using standard production vehicles with a prior art GNSS system and standard sensors with the accuracy typical of production vehicles to determine the position of the roadside structures. This allows for the cost-effective determination of the position of roadside structures. Furthermore, it enables the cost-effective and rapid detection of changes in the roadside structures. This makes it possible to create and maintain highly accurate roadside maps without the need for complex sensors. The roadside structures can be, for example, a guardrail along a highway or any other multi-lane road.The third values can be determined in a central unit, for example a so-called backend.
[0020] The third values of the object positions mfusion can be determined using a plurality of second values mv of the object positions (xi, yi), each determined by a plurality of first vehicles. The plurality of first vehicles may have been located at a satellite-based position within a tolerance band at the time the boundary development was detected. The accuracy of the method can be increased if the third values of the object positions mfusion are determined solely from probability distributions generated by vehicles that were located at approximately the same position when the boundary development was detected.
[0021] The procedure does not require the use of a stored model of a static object within the roadside structure. The object can be located next to the roadway and may include, for example, guardrails, signs, guideposts, or the like.
[0022] The step of determining the first probability distribution P(S|ZK) can include the step of determining a second probability distribution P(S|Zk), which indicates whether, at time k of the measurement, a boundary development exists at the object positions (xn, yn) determined by one of the first vehicles. The second probability distribution P(S|Zk) can be determined using a Gaussian process. The measurement at time k is given by Zk = {(xn, yn)|n = 1, ..., N} and includes all recorded first values of the position of the N objects of the boundary development at time k. The second probability distribution P(S|Zk) can make a statement about the probability of a boundary development being located at any given location. At locations where no measurement points (xn, yn) are nearby, the uncertainty is correspondingly higher than at locations where a measurement point (xn, yn) is nearby.
[0023] The step of determining the second probability distribution P(S|Z k), which checks whether there is a boundary structure at the object positions (xn , yn ) determined by a vehicle at time k, can include the step of generating a multi-modal probability distribution P(S|Z k ) over the boundary structure based on the plurality N of static objects detected at the object positions (xn , yn ) at time k. Based on successive measurements k = 1,..., K, the overall probability over the boundary structure at the current time K in the road plane P(S|ZK ) is obtained from all previously taken measurements at times k = 1, ..., K or at the individual times k = 1, ..., K-1 and the measurement at the current time K.
[0024] The step of determining the first probability distribution P(S|Z|K) can involve combining and / or predicting the successive measurements at times k = 1,..., K of the object positions in the road plane based on all previously determined second multi-modal probability distributions P(S|Zk) at times k = 1,..., K using the formula P S | Z K = 1 K ∑ k = 1 K P S | Z K carry out.
[0025] This step involves calculating the mean value using the probability distributions.
[0026] The procedure can determine a third probability distribution P(SO |ZK ) of the position of the static objects in a direction orthogonal to the direction of travel O from the first probability distribution P(S|ZK ) in the travel plane.
[0027] For positioning within the traffic lane, the orthogonal distance in the direction of travel or road to the roadside buildings is generally used. In the preceding steps, two-dimensional values were determined and could also be used for longitudinal positioning; however, this information is insufficient for longitudinal correction until corresponding landmarks, such as overhead sign gantries or bridges, are detected.
[0028] The method can determine a second value mv of the position of the edge buildings on the first side of the respective vehicle based on the third probability distribution P(SO |ZK ) of the static objects. The method can also determine an uncertainty σ v of the second value of the position of the edge buildings or the plurality of static objects on the first side of the respective first vehicle based on the third
[0029] Determine the probability distribution P(SO |ZK ) of the static objects. Each first vehicle can transmit the second value of the position of the majority of static objects on the first side and the uncertainty of the second value to a central unit. The uncertainty of the second value is essentially based on the first accuracy of the first value of the satellite-based position. The uncertainty can be a standard deviation or a variance. The first side can preferably be the left side.
[0030] The method can determine a second value for the position of the boundary structures on a second side of the respective first vehicle based on the third probability distribution P(SO |ZK ) of the static objects. Furthermore, the method can determine an uncertainty of the second value for the position of the majority of static objects on the second side of the respective first vehicle based on the third probability distribution P(SO |ZK ) of the static objects. The second value for the position of the majority of static objects on the second side and the uncertainty of the second value on the second side can be transferred to the central unit. The second side is opposite to the first side and is preferably the right side.
[0031] The procedure may further include the step of receiving the second position value and the uncertainty measure of the second position value from a plurality V of vehicles. The second value may be an absolute position or a relative position. If the second value is a relative position, the satellite-based position of the first vehicle in the plurality V of first vehicles may also be received.
[0032] The method can further determine a mean value mfusion of the position of the boundary structure and the position uncertainty σfusion of the boundary structure based on the second value of the position on the first and / or second side of the vehicle and the uncertainty of the value of the position on the first and / or second side of the vehicle. Generating the third values of the object positions mfusion and the position uncertainty σfusion can involve using a distance measure for an association problem for the relative distance to the detected boundary structure. For this purpose, the Mahalanobis distance, for example, can be used.
[0033] This serves to determine the values that are included in the fusion (for example, filtering outliers). If only relative coordinates are transmitted instead of global edge building coordinates, the determined distances of the edge buildings from the individual lanes must still be separated, for example, by the characteristic distance and, on German motorways, by an offset of approximately 3.5 m according to the standard, and / or merged with the individual estimates or probability distributions also from the same lane.
[0034] In this process, the same edge building is detected from multiple lanes at varying distances. Without information about the absolute position, several clusters of hypotheses for the distance to the edge building arise on a multi-lane roadway, i.e., for the identical edge building. Using the Mahalanobis distance, it can be determined which hypotheses create a representation of the multiple lanes (small Mahalanobis distance, uncertainty-weighted distance measure). The associated hypotheses are then merged.
[0035] Alternatively or additionally, the third values of the object positions m fusion of the positions of the perimeter buildings and / or the positional uncertainty σ fusion of the perimeter buildings can be calculated using the following formulas: m fusion = σ fusion 2 ∗ ∑ v = 1 V m v σ v 2 ; and σ fusion 2 = ∑ v = 1 V 1 σ v 2 − 1 ; where mv are the values for the position of the edge buildings in a direction orthogonal to the direction of travel and σv is the uncertainty for the position of the edge buildings in a direction orthogonal to the direction of travel based on the measurements of the vehicles v = 1, ..., V.
[0036] The fusion according to the above formula takes place at the respective local locations. The determined value m fusion and the uncertainty σ fusion, where the boundary development is located, apply locally to the location and / or the section of the route and / or to the GNSS position where this procedure was applied. Thus, for each desired GNSS coordinate, a fused position value or a third position value and an uncertainty according to the above formula result.
[0037] The method can further transmit the third values of the object positions (m fusion of positions) and / or the position uncertainty (σ fusion) to at least one vehicle. These third values and / or the position uncertainty can be transmitted to the vehicle based on its satellite-based position. The vehicle can be a first vehicle that has determined the object position values and the probability distributions. This provides the vehicle with a more accurate estimate of the surrounding buildings for a more precise determination of its own position. Alternatively, the vehicle can be a second vehicle that has not determined the positions of the surrounding buildings. Nevertheless, this vehicle receives an accurate estimate of the surrounding buildings for a more precise localization of its own position.
[0038] The procedure also includes the steps of correcting the satellite-based position determined by the first vehicle using the third values of the object positions m fusion as the likely position of the edge buildings, in order to generate corrected trajectories, and of determining clusters and / or a number of clusters of corrected trajectories, where the positions of the clusters each indicate a lane and the number of clusters indicates the number of lanes. The determination of the lanes and their number is therefore carried out based on the corrected trajectories of the vehicles.
[0039] The position of the roadside buildings, determined by the central unit, is transmitted to the vehicle. Consequently, the vehicle knows the position of the roadside buildings, for example, as the distance of the roadside buildings to a reference point or as the distance of the roadside buildings to the individual traffic lanes. The vehicle can now compare its distance measurement to the roadside buildings with the merged results or third-party values provided by the central unit. The vehicle can then position itself relative to the roadside buildings within the traffic lane, and no longer solely using GNSS data.
[0040] The invention also discloses a method for checking and / or changing the presumed position of a vehicle. The vehicle detects the position of the roadside buildings using a sensor. The position of the roadside buildings can be determined using the methods described above, for example, using the previously described first probability distribution P(S|ZK). The vehicle compares the position of the roadside buildings detected by the sensor with the expected position mfusion of the roadside buildings. The detected position of the roadside buildings can be determined using the previously described method. The expected position can be the previously determined third values of the object positions mfusion. The vehicle can be a first vehicle that has previously determined the values for the object positions of the roadside buildings. However, the vehicle can also be a second vehicle that has not previously determined any values for the object positions of the roadside buildings.
[0041] The invention also relates to a method for detecting a road construction site and / or for detecting a change in a road layout, which includes the steps described above.
[0042] The procedure for detecting a road construction site and / or a change in road alignment can check whether the determined probable position of the roadside buildings in a first area has changed within a predetermined period. Alternatively or additionally, the procedure for detecting a road construction site and / or a change in road alignment can check whether the probable position of the roadside buildings in the first area has changed based on the positions of the roadside buildings determined by at least a predetermined number of first vehicles. The predetermined number can include a frequency, for example, a predetermined number within a predetermined period.If the determined probable position of the edge of the area changes within a predetermined period in the first zone, the method can indicate that a construction site may be present in the first zone and / or that the road alignment has changed. The first zone can be linked to the GNSS position to allow it to be assigned to other vehicles. The predetermined period can be at least 30 minutes. This allows for the detection of mobile construction sites and / or daytime construction sites. Preferably, the predetermined period can be at least 2 hours, and most preferably at least one day. This provides a larger basis for statistical analysis. The predetermined number of vehicles can be at least 5, preferably 10, and more preferably at least 25.
[0043] The procedure can check whether multiple candidates are identified as the likely positions of the road's edge in a second area. If multiple candidates are identified as the likely positions of the road's edge in the second area, the procedure can indicate that there may be a construction site in the second area and / or that the road alignment has changed. These steps can detect that a road alignment has been divided. A statistical assumption and / or a frequency distribution can be used for this purpose. In contrast to the detection of the position and / or number of lanes, the step of detecting multiple candidates does not use the trajectories whose GNSS position has been corrected, but rather the initial probability distributions determined by the individual vehicles.
[0044] The invention also relates to a method for detecting the perimeter buildings of a road construction site, which comprises the steps described above.
[0045] The invention also relates to a computer program product which, when loaded into a memory of a computer with a processor, performs the steps of one of the previously described methods.
[0046] The invention will now be described with reference to the accompanying figures, which show non-limiting embodiments of the invention, wherein Figure 1 a plurality of trajectories are shown, as they are captured by a satellite-based positioning system; Figure 2 Objects of the perimeter development detected by a radar; Figure 3 Positions of objects in the perimeter development shown after filtering; Figure 4 Data of the perimeter development shows how it was recorded by a vehicle; Figure 5Data from the perimeter development, recorded by several vehicles, is shown; Figure 6 shows the likely position of the perimeter development; Figure 7 a scenario without a construction site; Figure 8 the determination of the vehicle's position relative to the surrounding buildings shows; Figure 9 shows lane positions; and Figure 10 a scenario with a construction site is shown.
[0047] Figure 1Figure 1 shows a two-lane motorway 100, where carriageway 101 has three lanes on which vehicles can travel. The vehicles determine their position on carriageway 101 using satellite-based positioning systems. Such satellite-based positioning systems, for example GPS, have an accuracy of a few meters. This level of accuracy is insufficient to determine which lane of carriageway 101 the vehicle is traveling in. Lines 102 show the vehicles' own positions as determined by the satellite-based system. Lines 102 extend across the entire width of carriageway 101.
[0048] As previously described, an important requirement for the use of autonomous vehicles is that the position of the vehicle can be determined so accurately that the position of the vehicle can be assigned to a lane.
[0049] It will be on Figure 2Reference is made to the first line 102, which shows the trajectories of a vehicle and the positions of static objects Z k 106, 108, 110, 112, which were determined by a sensor in the vehicle, for example, by a radar sensor and / or a lidar sensor. The first line 102 shows the trajectory of the vehicle that determined the object positions 106, 108, 110, 112.
[0050] The second line 104 shows the mean trajectory of all in Figure 1 trajectories shown 102.
[0051] The vehicle on trajectory 102 determines the positions Zk for objects that could potentially be part of a boundary structure, such as a guardrail. These positions are designated by reference symbols 106, 108, 110, and 112. For each of the N objects at positions (xn, yn), n = 1, ..., N, a plurality of initial position values are determined at times k = 1, ..., K, but not necessarily for every time k. Due to the accuracy of the vehicle's satellite-based position measurement, the positions of the objects forming the boundary structure are not determined precisely, but with a measurement error on the order of a few meters.
[0052] It will be on Figure 3Reference is made to a graph whose abscissa represents the distance the vehicle travels along trajectory 102, and whose ordinate represents the measured position of the edge structure. In the described embodiment, the method according to the invention uses the positions 106 of the edge structure on the left side of the vehicle that are within a predetermined tolerance band. Consequently, measured positions 108 that are outside the tolerance band on the left side are deleted. Since in the Figure 3 In the considered embodiment, since the edge structure on the right side of the vehicle is not relevant, the measured object positions 110 and the scatter values 112 below the abscissa are deleted or not evaluated further. The mean trajectory 104 forms the abscissa as a reference.
[0053] It will be on Figure 4Reference is made to point 106, which shows the recorded positions of objects in the perimeter development as captured by a single vehicle. Line 120 shows the mean of the probability distribution P(SO |ZK ) as the second value of the measured positions of the perimeter development. Each object in the perimeter development is measured multiple times as the vehicle travels along the perimeter, since the sensor, for example, the radar sensor, captures the vehicle's surroundings across a wide angular range, for example, due to the sensor's placement at the front of the vehicle and / or its range.
[0054] Line 122 shows the dispersion of the mean, for example, the variance. The mean trajectory 104 forms the abscissa as a reference.
[0055] The mean does not necessarily have to be an arithmetic or geometric mean. The mean can be determined using a statistical method, for example a so-called Gaussian process, which is known to those skilled in the art.
[0056] Line 124 shows the deviation of the vehicle's trajectory 102, which determined the positions 106 of the objects in the perimeter development, from the mean trajectory 104. Line 124 thus represents trajectory 102 with the mean trajectory 104 as a reference.
[0057] It will be on Figures 5 , 6 and 7Reference is made to the figures shown, which demonstrate how the probable position as the third value P(SO |ZK ) of the position of objects of the perimeter development is determined using the second values determined by a plurality of motor vehicles. The set of lines 125 shows the presumed positions (second values) of the perimeter development 138, as determined by individual vehicles 132, 134, 136. Each line 120 is generated by one vehicle. Each line 120 forms a part of the set 125. Line 138 shows the probable position of the left perimeter development, as determined by the method according to the invention, for example in a central device (backend). The trajectories 132, 134, 136 are three trajectories from the set of trajectories 102 of Figure 1 and were corrected in their position according to the procedure.
[0058] The likely position 138 (third value) of the left-hand edge development can be determined, for example, using the following formulas: m fusion = σ fusion 2 ∗ ∑ v = 1 V m v σ v 2 ; and σ fusion 2 = ∑ v = 1 V 1 σ v 2 − 1 ; where mv are the values for the position of the edge buildings in a direction orthogonal to the direction of travel and σv is the uncertainty for the position of the edge buildings in a direction orthogonal to the direction of travel based on the measurements of the vehicles v = 1, ..., V.
[0059] The likely position of the roadside buildings can also be determined using the Mahalanobis distance. This serves to determine the values included in the fusion (for example, filtering outliers). If only relative coordinates are transmitted instead of global roadside building coordinates, the determined distances of the roadside buildings from the individual lanes must be separated, for example, by the characteristic distance and, on German highways, by an offset of approximately 3.5 m according to the standard, and / or merged with the individual estimates from the same lane.
[0060] The procedure therefore determines the likely number of lanes and the likely distance of 130 lanes to the left-hand edge of the road. This data is transmitted to vehicles traveling on the corresponding section of road.
[0061] Subsequently, with reference to Figures 4 and 6 to 8It describes how a vehicle 142 estimates its own position relative to the roadside buildings. For a vehicle 142, the reference trajectory is 104 (probable position of the lane), the own trajectory is 102, and the own roadside building estimate is 120, which is determined as previously described with reference to... Figures 1 to 4 as described, and the general perimeter development estimate 130 based on several vehicles, which is determined as previously described with reference to Figures 5 to 7 As described, line 130 indicates the probable distance D of the edge buildings from the reference trajectory 104 of the middle lane. Furthermore, it could define where the (three) lanes are located in relation to the reference trajectory 104.
[0062] Figure 8The diagram shows a three-lane highway. Trajectory 104a shows the likely center position of the right lane, trajectory 104 shows the likely center position of the middle lane, and trajectory 104b shows the likely center position of the left lane. Trajectory 102a shows the actual trajectory traveled by a vehicle in the right lane, trajectory 102 shows the actual trajectory traveled by a vehicle in the middle lane, and trajectory 102b shows the actual trajectory traveled by a vehicle in the left lane.
[0063] The vehicle 142 determines its own distance dv from its own trajectory 102 to the edge development 138 using its own edge development estimate 120 by comparison with the general edge development estimate 130.
[0064] Assuming that the roadside buildings are generally identical to those in real life, the vehicle can position itself based on these buildings and significantly reduce inaccuracies using GNSS-based data. The algorithm used for this purpose can, for example, be based on the SLAM (Simultaneous Localization and Mapping) method.
[0065] With reference to Figures 8 and 9 The determination of the position and / or number of lanes is explained. The trajectories 102, as they appear in Figure 1The trajectories shown will be corrected using the previously described procedure so that the positions of trajectories 103, 103a, and 103b are relative to the edge building 138. Each vehicle compares its trajectory and the determined distance from the edge building with the average distance from the edge building and the reference trajectory transmitted by the central unit. Each vehicle then corrects its own trajectory accordingly with the value from the comparison, resulting in trajectories 103, 103a, and 103b. The vehicle may use knowledge of the lane spacing and / or the lane distance from the edge building in this process.
[0066] Trajectories 103 show the actual trajectories of vehicles that traveled in the center lane. Trajectories 103a show the actual trajectories of vehicles that traveled in the left lane. Trajectories 103b show the actual trajectories of vehicles that traveled in the right lane. By identifying clusters, the number and / or position of lanes can be determined. A lane can be considered a cluster in the central unit if the number of vehicles whose trajectory belongs to a cluster exceeds a predetermined threshold. This allows, for example, trajectories 105 of vehicles changing lanes to be filtered out, since only a few vehicles (whose number is lower than the threshold) change lanes within a given area.
[0067] In Figure 10A scenario is shown that takes place in the same location as the one in Figures 1 to 6 The scenario shown is located in the Figure 10 The scenario shown is a construction site on the Figure 8 The section of the motorway shown was set up. Parts of the roadway were closed for the construction site. Consequently, other lanes had to be set up. Vehicles 160, 162, and 164, traveling on a first separated lane in the area of a trajectory 152, determine a probable distance from the first edge building 170. The probable distance of the edge building determined by vehicles 160, 162, and 164 yields a first set 154 of probable positions of the edge building. Vehicles 166 and 168, traveling on a second separated lane in the area of a trajectory 150, which is structurally separated from the first lane, determine a second set 156 of probable positions of the edge building 172.
[0068] The central institution determines, firstly, that the position of the perimeter buildings is in relation to the one in Figures 1 to 6 The scenario shown has changed within a predetermined timeframe. The number in the Figure 10 In the scenario shown, the changed probable positions of the perimeter buildings exceed a predetermined threshold. Therefore, the procedure can determine that the perimeter buildings have been changed due to a construction site or a structural alteration.
[0069] Secondly, the central authority can determine that, based on the two sets 154 and 156, there are several candidates for the likely position of the perimeter building. Therefore, the central authority can determine that the perimeter building has been altered due to a construction site or a structural change.
[0070] The method according to the invention enables the creation and updating of navigation data with high accuracy without the use of specialized measuring vehicles. Furthermore, the method according to the invention can be used for the detection of construction sites.
Claims
1. Method for estimating the position of the peripheral development (138) pertaining to a road, comprising the following steps: - travelling along a road section with a plurality of first vehicles each moving along a trajectory (102), each of the plurality of first vehicles performing the following steps: - determining a satellite-based position of the respective first vehicle by means of a satellite-based positioning system using position determination means of the respective first vehicle when it moves along the trajectory (102); and - determining a position of the peripheral development by means of a measurement using a sensor of the first vehicle and on the basis of the satellite-based position of the first vehicle; and - sending the determined position of the peripheral development to a central device, a probable position of the peripheral development being determined by virtue of the following steps being performed by the central device: - receiving a plurality of determined positions of the peripheral development, each of which has been determined by one of the plurality of first vehicles, at the central device; and - determining the probable position of the peripheral development on the basis of the plurality of determined positions of the peripheral development that have each been determined by one of the plurality of first vehicles, characterized by the following steps: - determining first values of object positions (106 , 108 , 110 , 112) Zk = {(xn, yn) |n = 1, ..., N}, as a determined position of the peripheral development at the time k, pertaining to a plurality N of static objects Zk outside the roadway with respect to the respective first vehicle by means of the sensor of the respective first vehicle at a plurality of times k = 1, ..., K and on the basis of the satellite-based position of the first vehicle; - sending the first values to the central device as a determined position of the peripheral development; - determining a first probability distribution P(S|ZK), which indicates whether there is a peripheral development at second values mv (120) of positions (xi, yi), on the basis of the plurality K of determinations of the first values of object positions ZK at the times k for k = 1, ... K, where P(S|ZK) represents the conditional probability of a static peripheral development S and where ZK comprises all past determinations, at the times k, of first values of the object positions Zk and the current determination, at the time K, of first values of the object positions ZK of the first vehicle; - sending the first probability distribution to the central device as a determined position of the peripheral development; - receiving a plurality of first probability distributions for the probable position of the peripheral development, each of which has been determined by one of the plurality of first vehicles, at the central device as a determined position of the peripheral development; and - determining third values (130) of the object positions mfusion as a probable position of the peripheral development on the basis of the plurality of first probability distributions (120) that have each been determined by the plurality of different first vehicles.
2. Method according to Claim 1, characterized in that the third values (130) of the object positions mfusion are determined by means of the plurality of second values mv (120) that have each been determined by the plurality of first vehicles.
3. Method according to Claim 1 or 2, characterized in that the step of determining a first probability distribution (120) P(S|ZK) comprises the following step: - determining a second probability distribution P(S|Zk) that indicates whether, at the time k of the measurement, there is a peripheral development at the object positions (xn, yn) determined by one of the first vehicles.
4. Method according to Claim 3, characterized in that the step of determining the second probability distribution P(S|Zk) regarding whether, at the time of the measurement k, there is a peripheral development at the object positions (xn, yn) determined by one of the first vehicles comprises the following step: - generating a first multimodal probability distribution P(S|Zk) relating to the peripheral development on the basis of the plurality N of static objects (106) identified at the object positions (xn, yn) at the time k.
5. Method according to Claim 4, characterized by the following step: - determining a second multimodal probability distribution P(S|ZK) on the basis of the first multimodal probability distributions P(S|ZK) pertaining to the first value of the object positions (xn, yn) of the at least one static object during chronologically successive measurements at the times k = 1, ..., K.
6. Method according to Claim 5, characterized in that the step of determining the second multimodal probability distribution P(S|ZK) comprises the following step: - merging and / or predicting the chronologically successive measurements at the times k = 1,..., K of the object positions on the driving plane on the basis of all previously determined first probability distributions P(S|ZK) at the individual times k by means of the formula: P S | Z K | K = 1 K ∑ k = 1 K P S | Z k 7. Method according to one of the preceding claims, characterized in that the at least one static object is detected without a stored model.
8. Method according to one of the preceding claims, characterized by the step of determining a third probability distribution P(SO|ZK) pertaining to the position of the static objects (106) in a direction orthogonal to the direction of travel O from the first probability distribution P(S|ZK) on the driving plane.
9. Method according to Claim 8, characterized by the following step: - determining a second value mv of the position (120) of the peripheral development on a first side of the respective first vehicle on the basis of the third probability distribution P(SO|ZK) pertaining to the static objects (106); - determining an uncertainty σv (122) pertaining to the second value of the position of the peripheral development on a first side of the respective first vehicle on the basis of the third probability distribution P(SO|ZK) pertaining to the static objects (106); and - transmitting the second value mv of the position (120) of the peripheral development on the first side and the uncertainty (122) pertaining to the second value to a central unit.
10. Method according to Claim 9, characterized by the following steps: - determining a second value mv of the position (120) of the peripheral development on a second side of the respective first vehicle on the basis of the third probability distribution P(SO|ZK) pertaining to the static objects (106); - determining an uncertainty σv (122) pertaining to the second value of the position on the second side of the respective first vehicle on the basis of the third probability distribution P(SO|ZK) pertaining to the plurality of static objects; and - transmitting the second value of the position (120) of the plurality of static objects on the second side and the uncertainty (122) pertaining to the second value to the central unit.
11. Method according to either of Claims 9 and 10, characterized by the following steps: - receiving the second value mv of the position (120) of the peripheral development and the measure of uncertainty σv (122) pertaining to the second value (120) of the position of the peripheral development from a plurality V of first vehicles (102); and - determining the position mfusion (130) of the peripheral development (138) and the positional uncertainty σfusion pertaining to the peripheral development (138) on the basis of the second value of the position (120) and the uncertainty pertaining to the second value of the position (122) on the first and / or second side of the vehicle.
12. Method according to one of Claims 9 to 11, characterized in that the step of generating the third values (130) of the positions mfusion of the peripheral development and the positional uncertainty σfusion pertaining to the peripheral development comprises at least one of the following: - using a measure of distance; and / or - computing the third values (130) of the object positions mfusion pertaining to the positions of the peripheral development (106) and the positional uncertainty σfusion pertaining to the peripheral development by using the following formulae: and und 13. Method according to Claim 11 or 12, characterized by the following step: - transmitting the position of the peripheral development (138) mfusion and the positional uncertainty σfusion pertaining to the peripheral development (138) to at least one vehicle.
14. Method according to one of Claims 11 to 13, characterized by the following steps: - correcting the satellite-based position determined by the respective first vehicle by means of the third values (130) of the object positions mfusion as a probable position of the peripheral development, in order to generate corrected trajectories (103 , 103a , 103b); and - determining accumulations and / or a number of accumulations of corrected trajectories (103 , 103a , 103b), the positions of the accumulations indicating one lane each and the number of accumulations indicating a number of lanes.
15. Method for checking and / or changing the presumed position of a vehicle, characterized by the following steps: - detecting a detected position of the peripheral development (106) by means of a vehicle travelling on a road; and - comparing the position of the peripheral development as detected by the vehicle with an expected position of the peripheral development (130); the detected position of the peripheral development (106) being determined by means of the method according to one of Claims 1 to 14.
16. Method for identifying roadworks and / or for identifying a change of road layout, comprising the steps of the method according to one of Claims 1 to 14.
17. Method according to Claim 16, characterized by the following steps: - checking whether the determined probable position of the peripheral development in a first region has changed within a predetermined period of time and / or whether, on the basis of the positions of the peripheral developments as determined by at least a predetermined number of first vehicles, the probable position of the peripheral development in the first region has changed; and - if the determined probable position of the peripheral development in the first region has changed, indicating that there may be a construction site, and / or the road layout has changed, in the first region.
18. Method according to Claim 16 or 17, characterized by the following steps: - checking whether a plurality of candidates are determined as a determined probable position of the peripheral development in a second region; and - if a plurality of candidates are determined as a determined probable position of the peripheral development in the second region, indicating that there may be a construction site, and / or the road layout has changed, in the second region.
