Method and system for mapping an environment

Abstract

Method (800) for mapping an environment, comprising: Calculating a multitude of radial components and a multitude of angular components for a positive inverse sensor model (ISM) of an occupancy grid (200, 400, 600); Receiving a detection (204, 212) at a sensor (202) of an object in an environment surrounding a vehicle (10); Selecting a radial component corresponding to a detection area from the multitude of radial components and selecting an angular component corresponding to a detection angle from the multitude of angular components; Multiplying the selected radial component and the selected angular component to generate an occupancy grid (200, 400, 600) for detection; and Mapping the surroundings using the occupancy grid (200, 400, 600).

Description

[0001] The subject matter of the disclosure relates to methods for mapping an environment using one or more radar systems and, in particular, to the use of a negative inverse sensor model that can be updated via a quad-tree structure to form an occupancy grid for representing the environment.

[0002] Radar systems can be used on vehicles to detect objects in the vehicle's surroundings for environmental mapping purposes or to detect objects within the environment. Such maps can be used to navigate the vehicle through the environment without making contact with objects. One method for processing radar signals involves creating a volumetric occupancy grid that captures object locations in the form of probabilities based on radar system detections. The occupancy grid can be created and updated using a dual inverse sensor model (ISM), which includes a positive ISM and a negative ISM. An important aspect of mapping an environment using an occupancy grid is the ability to update the grid to reflect the current state of the environment and the objects within it.

[0003] WERBER, Klaudius [et al.]: Automotive radar gridmap representations describes representations of automotive radar grid maps. In robotics applications, grid maps are a common representation of the environment. Radar is well-suited as a sensor technology for the automotive sector due to its robustness. This article presents two radar-based grid mapping algorithms for automotive applications such as self-localization. These algorithms include, firstly, an amplitude-based approach that obtains information about the radar coordinate system (RCS) of all targets, and secondly, an occupancy grid mapping approach with a modified inverse sensor measurement model. Experiments show that both grid mapping algorithms provide an adequate representation of the environment.

[0004] Accordingly, the object of the present invention is to provide a method for imaging the environment that updates at least the negative ISM quickly and efficiently.

[0005] The problem is solved by the subject matter of the independent claims.

[0006] A method according to the invention includes calculating a plurality of radial components and a plurality of angular components for a positive inverse sensor model (ISM) of an occupancy grid, obtaining a detection at a sensor of an object in an environment surrounding the vehicle, selecting a radial component corresponding to an area of ​​detection from the plurality of radial components, and selecting an angular component corresponding to an angle of detection from the plurality of angular components, multiplying the selected radial component and the selected angular component to generate an occupancy grid for detection, and mapping the environment using the occupancy grid.

[0007] In addition to one or more of the features described herein, the ISM further comprises a negative ISM, and a probability value of the negative ISM is assigned using a quad-tree structure of the occupancy grid. A node of the quad-tree structure corresponds to a solid angle, and a first level of the quad-tree structure contains solid angle storage compartments that are combined to cover a field of view of the sensor. The method further comprises dividing a solid angle storage compartment on a level of the quad-tree structure into a plurality of solid angles on a level of the quad-tree structure if the solid angle storage compartment on that level contains the detection. The method further comprises assigning a probability value to a solid angle storage compartment for which there is no detection, the probability value reflecting the absence of detection in the solid angle storage compartment.The method further involves combining the probability values ​​from the negative ISM with a probability value from the positive ISM to create the occupancy grid. In various embodiments, the method includes navigating the vehicle relative to the object based on mapping the environment.

[0008] Furthermore, a system according to the invention for mapping the environment of a vehicle is disclosed. The system comprises a sensor and a processor. The sensor is configured to detect an object in the environment surrounding the vehicle. The processor is configured to calculate a plurality of radial components and a plurality of angular components for a positive inverse sensor model (ISM) of an occupancy grid, to select a radial component corresponding to a detection area from the plurality of radial components and an angular component corresponding to a detection angle from the plurality of angular components, and to multiply the selected radial component and the selected angular component to generate an occupancy grid for detection and to map the environment using the occupancy grid.

[0009] In addition to one or more of the features described herein, the processor is further configured to assign a probability value to a negative ISM of the occupancy grid using a quad-tree structure of the occupancy grid. A node of the quad-tree structure corresponds to a solid angle, and a first level of the quad-tree structure contains solid angle storage compartments that are combined to cover a field of view of the sensor. The processor is further configured to partition a solid angle storage compartment on a level of the quad-tree structure into a plurality of solid angles on a level of the quad-tree structure if the solid angle storage compartment on that level contains the detection. The processor is further configured to assign a probability value to a solid angle storage compartment for which there is no detection, the probability value reflecting the absence of detection in the solid angle storage compartment.The processor is further configured to combine the probability values ​​from the negative ISM with a probability value from a positive ISM to create the occupancy grid. In various embodiments, the processor is configured to navigate the vehicle relative to the object based on mapping the environment.

[0010] In an application example, a vehicle with a system according to the invention is disclosed. The vehicle includes a sensor and a processor. The sensor is configured to detect an object in an environment surrounding the vehicle. The processor is configured to calculate a plurality of radial components and a plurality of angular components for a positive inverse sensor model (ISM) of an occupancy grid, to select a radial component corresponding to a detection area from the plurality of radial components and an angular component corresponding to a detection angle from the plurality of angular components, and to multiply the selected radial component and the selected angular component to generate an occupancy grid for detection and to map the environment using the occupancy grid.

[0011] In addition to one or more of the features described herein, the processor is further configured to assign a probability value to a negative ISM of the occupancy grid using a quad-tree structure of the occupancy grid. A node of the quad-tree structure corresponds to a solid angle, and a first level of the quad-tree structure contains solid angle storage compartments that are combined to cover a field of view of the sensor. The processor is further configured to partition a solid angle storage compartment on a level of the quad-tree structure into a plurality of solid angles on a level of the quad-tree structure if the solid angle storage compartment on that level contains the detection. The processor is further configured to assign a probability value to a solid angle storage compartment for which there is no detection, the probability value reflecting the absence of detection in the solid angle storage compartment.The processor is further designed to combine the probability values ​​from the negative ISM with a probability value from a positive ISM to create the occupancy grid.

[0012] Other features, advantages and details appear, only by way of example, in the following detailed description of the embodiments, the detailed description referring to the drawings, wherein the following applies: Fig. Figure 1 shows a vehicle with an associated trajectory planning system, which is depicted at 100 according to various embodiments; Fig. Figure 2 shows an illustrative occupancy grid; Fig. Figure 3A (state of the art) shows a radial probability diagram for creating or updating an occupancy grid based on the position of an object; Fig. Figure 3B (state of the art) shows a log(ratio) model for a radial probability diagram of an object located within a selected range; Fig. 4 shows an occupancy grid illustrating areas without detection; Fig. Figure 5 shows an illustrative quad-tree structure that can be used to store and update probability values ​​for a negative ISM; Fig. Figure 6 illustrates the quad-tree structure of Fig. 5, to fill an occupancy grid using a negative ISM; Fig. Figure 7 shows a positive ISM, a negative ISM, and a dual ISM for an occupancy grid; Fig. Figure 8 shows a flowchart illustrating a procedure for generating an occupancy grid that represents an environment of the vehicle. Fig. 1 dynamically displays; Fig. Figure 9 shows a representation of an illustrative radar field in a scenario where an object passes in front of the vehicle, the radar field being formed from an occupancy grid that does not use the negative ISM; and Fig. Figure 10 shows the representation of the illustrative radar field in the same scenario as Fig. 9, wherein the radar field is formed from an occupancy grid that uses the negative ISM disclosed herein.

[0013] According to an exemplary embodiment, Fig. 1. A vehicle 10 with an associated trajectory planning system, which is shown at 100 according to various embodiments. In general, the trajectory planning system 100 determines a trajectory plan for the automated driving of the vehicle 10. The vehicle 10 generally comprises a chassis 12, a body 14, front wheels 16, and rear wheels 18. The body 14 is arranged on the chassis 12 and essentially encloses the components of the vehicle 10. The body 14 and the chassis 12 can together form a frame. The wheels 16 and 18 are each rotatably coupled to the chassis 12 near a respective corner of the body 14.

[0014] In various embodiments, the vehicle 10 is an autonomous vehicle, and the trajectory planning system 100 is integrated into the autonomous vehicle 10 (hereinafter referred to as the autonomous vehicle 10). The autonomous vehicle 10 is, for example, a vehicle that is automatically controlled to transport passengers from one place to another. The autonomous vehicle 10 is depicted as a passenger car in the illustrated embodiment; however, it should be noted that any other vehicle, including motorcycles, trucks, sports vehicles (SUVs), recreational vehicles (RVs), watercraft, aircraft, etc., can also be used. In one exemplary embodiment, the autonomous vehicle 10 is a so-called Level Four or Level Five automation system.A Level Four system indicates a "high level of automation" by referring to the driving mode-specific performance of an automated driving system in all aspects of the dynamic driving task, even if a human driver is unable to intervene appropriately. A Level Five system indicates "full automation" and refers to the full-time performance of an automated driving system in all aspects of the dynamic driving task under all road and environmental conditions that can be managed by a human driver.

[0015] As shown, the autonomous vehicle 10 generally includes a drive system 20, a transmission system 22, a steering system 24, a braking system 26, a sensor system 28, an actuator system 30, at least one data storage device 32, and at least one controller 34. The drive system 20 may, in various embodiments, include an internal combustion engine, an electric machine such as a traction motor, and / or a fuel cell drive system. The transmission system 22 is configured to transmit power from the drive system 20 to the vehicle wheels 16 and 18 according to selectable gear ratios. According to various embodiments, the transmission system 22 may include a gear-ratio automatic transmission, a continuously variable transmission, or another suitable transmission. The braking system 26 is configured to provide braking torque to the vehicle wheels 16 and 18.The braking system 26 can, in various embodiments, include friction brakes, brake-by-wire, a regenerative braking system such as an electric motor, and / or other suitable braking systems. The steering system 24 influences the position of the vehicle wheels 16 and 18. While in some embodiments within the scope of this disclosure it is depicted as a steering wheel for illustrative purposes, the steering system 24 may not include a steering wheel.

[0016] The sensor system 28 includes one or more sensor devices 40a-40n that detect observable states of the external and / or internal environment of the autonomous vehicle 10. The sensor devices 40a-40n may include, but are not limited to, radar systems, lidar, global positioning systems, optical cameras, thermal imaging cameras, ultrasonic sensors, and / or other sensors. In various embodiments, the vehicle 10 includes a radar system with an array of radar sensors, the radar sensors being located at various points along the vehicle 10. During operation, a radar sensor emits an electromagnetic pulse 48, which is reflected on the vehicle 10 by one or more objects 50 within the sensor's field of view. The reflected pulse 52 appears as one or more detections at the radar sensor.

[0017] The actuator system 30 includes one or more actuator devices 42a-42n that control, but are not limited to, one or more vehicle characteristics, such as the drive system 20, the transmission system 22, the steering system 24, and the braking system 26. In various embodiments, the vehicle characteristics may also include, but are not limited to, interior and / or exterior vehicle features, such as doors, a trunk, and interior features such as ventilation, music, lighting, etc. (not numbered).

[0018] The controller 34 includes at least one processor 44 and a computer-readable memory device or media 46. The processor 44 can be a custom-designed or commercially available processor, a central processing unit (CPU), a graphics processing unit (GPU) among multiple processors connected to the controller 34, a semiconductor-based microprocessor (in the form of a microchip or chipset), a macroprocessor, a combination thereof, or generally any device for executing instructions. The computer-readable memory device or media 46 can include volatile and non-volatile memory in the form of read-only memory (ROM), direct-access memory (RAM), and keep-alive memory (KAM). KAM is persistent or non-volatile memory that can be used to store various operating variables while the processor 44 is powered off.The computer-readable storage device or media 46 can be implemented using any number of known storage devices, such as PROMs (programmable read-only memory), EPROMs (electrical PROMs), EEPROMs (electrically erasable PROMs), flash memory, or any other electrical, magnetic, optical, or combined storage devices capable of storing data, some of which may be executable instructions used by the controller 34 in controlling the autonomous vehicle 10.

[0019] The instructions can include one or more separate programs, each containing an ordered list of executable instructions for implementing logical functions. When executed by the processor 44, the instructions receive and process signals from the sensor system 28, perform logic, calculations, procedures, and / or algorithms for the automatic control of the components of the autonomous vehicle 10, and generate control signals to the actuator system 30 to automatically control the components of the autonomous vehicle 10 based on the logic, calculations, procedures, and / or algorithms. Although in Fig. 1 where only one controller 34 is shown, embodiments of the autonomous vehicle 10 may include any number of controllers 34 which communicate and interact via a suitable communication medium or a combination of communication media to process the sensor signals, perform logics, calculations, procedures and / or algorithms, and generate control signals to automatically control the functions of the autonomous vehicle 10.

[0020] The trajectory planning system 100 navigates the autonomous vehicle 10 based on the identification of objects and / or their locations within the vehicle's environment. In various embodiments, the controller 34 performs calculations to record probabilities in an occupancy grid based on detections from the vehicle 10's radar system. The processor uses the occupancy grid to make decisions regarding the vehicle 10's navigation. After determining various object parameters, such as range, azimuth, elevation, speed, etc., from the occupancy grid, the controller 34 can operate one or more actuators 42a-n, the drive system 20, the transmission system 22, the steering system 24, and / or the brake 26 to steer the vehicle 10 relative to the object 50. In various embodiments, the controller 34 navigates the vehicle 10 to avoid contact with the object 50.

[0021] In one aspect, the processor 44 generates an occupancy grid using a dual inverse sensor model (ISM). An occupancy grid is a three-dimensional representation of the vehicle's environment 10. The occupancy grid represents a volume of the environment that can be detected using the radar system and is therefore limited in range, azimuth, and elevation. The occupancy grid is described using spherical polar coordinates (ρ, θ, φ) centered on the sensor. The occupancy grid is divided into three-dimensional storage compartments, labeled by (ρ, θ, φ), and stores or records probabilities for radar detections in compartments corresponding to the detection range (ρ), azimuth (θ), and elevation (φ). When navigating a vehicle, the values ​​of the occupancy grid are retrieved to ensure a probabilistic representation of the environment and the objects within it.The values ​​within the occupancy grid can thus be used to inform processor 44 about objects, so that processor 44 can perform operations on vehicle 10 to prevent vehicle 10 from making contact with the objects.

[0022] The occupancy grid is created and updated using a dual inverse sensor model (ISM), which includes a positive inverse sensor model and a negative inverse sensor model. An inverse sensor model is a model that specifies a distribution of state variables (i.e., in the occupancy grid) based on measurements. The positive ISM assigns a conditional occupancy probability to a storage location in the occupancy grid based on a measurement indicating the presence of a detection at a corresponding location in the environment. The positive ISM uses probabilities to represent the presence of the object at a specific location in the environment. The negative ISM tracks areas where no detections occur and assigns appropriate probabilities to the corresponding storage locations.In one embodiment, the positive ISM and the negative ISM are updated separately using radar observations, wherein the radar observation includes areas or directions where detections occur and areas or directions where no detections occur. The areas or directions where no directions occur (non-detection directions) are used to update the negative ISM, and the areas or directions where directions are found (with-detection directions) are used to update the positive ISM. The negative ISM and the positive ISM are then combined to create an updated dual ISM. The updated dual ISM is used to create a static occupancy grid.

[0023] To derive an ISM, one begins with a probabilistic description of the sensor, i.e., a distribution of sensor readings for a target at a specific location X. Bayes' theorem is applied to obtain the probability that the target is located at location X for a given sensor reading. To create the occupancy grid, the representation of the space surrounding the vehicle 10, as shown in Fig. 2 shown, discretized by introducing a three-dimensional grid.

[0024] The grid cells of the three-dimensional grid are considered independent of each other, so their occupancy probabilities correspond to the following equation: p(m1,m2,…,mN|X)=∏k=1Np(mk|X) where m k = 0 or 1. An empty k th Grid point has m k = 0 and an occupied k th Grid point has m k = 1.

[0025] Fig. Figure 2 shows an illustrative occupancy grid 200. A spherical polar coordinate system 225 is centered on a sensor 202 of a radar system. The sensor 202 is located at the origin of the spherical polar coordinate system 225. For illustration, a first detection 204 and a second detection 212 are shown. For range variables, the first detection 204 lies within a specific range storage compartment. For angular variables, the first detection 204 lies within an azimuthal storage compartment 206 and an elevation storage compartment 208. The azimuthal storage compartment 206 and the elevation storage compartment 208 define a first solid angle storage compartment 210, which covers an area on a spherical surface centered on the sensor 202.The first solid angle storage compartment 210 can be determined by projecting a beam from sensor 202 through the first detection 204 onto the spherical surface and recording the storage compartment through which the beam passes. Similarly, the second detection 212 lies within an azimuthal storage compartment 214 and an elevation storage compartment 216. The azimuthal storage compartment 214 and the elevation storage compartment 216 define a second solid angle storage compartment 218 of the spherical surface. The second solid angle storage compartment 218 can be determined by projecting a beam from sensor 202 through the second detection 212 onto the spherical surface and recording the storage compartment through which the beam passes.

[0026] In various embodiments, it is useful to use a logarithm of probability (i.e., the "odds") to work with log(probability) or "log(odds)." An odds ratio is defined as the ratio between the probability that a cell is occupied and the probability that the cell is empty, which is given by Eq. (2): Ok=p(mk=1|X)p(mk=0|X)=p(mk=1|X)1−p(mk=1|X)

[0027] Using the logarithm of the odds, O generates k a log(quotient) value L k : Lk=log Ok=log[p(mk=1|X)1−p(mk=1|X)]

[0028] The log(ratio) values ​​simplify the process of updating the ISMs because they are additive in the logarithm domain. They can also be calculated recursively for any iteration n by simple addition: Lkn=Lkn−1+ΔLkn−Lk0 where ΔLkn The change in the logarithmic probability at cell k after measurements at time n is. ΔLk0 is often, but not necessarily, considered to be zero.

[0029] The positive ISM is used to ΔLkn To estimate the occupancy probabilities of various locations with detection storage compartments in the occupancy grid, it is used to estimate the occupancy probabilities at the detection storage compartments of the occupancy grid at a selected time. Due to the high angular resolution, high detection probability, and low false alarm rates, positive ISM models are represented by delta functions along the azimuth and elevation.

[0030] For the positive ISM, the spatial tolerances in the detections are due to variations. (σρ2,σθ2,σφ2) The tolerances are characterized by range, azimuth, and elevation, respectively. These tolerances are considered uncorrelated. In one embodiment, the probability density function for detection is approximately Gaussian, as in Eq. (5): p(ρ,θ,φ)~N(ρ−ρ0,σρ2)N(θ−θ0,σθ2)N(φ−φ0,σφ2)

[0031] A recognition i at location (ρ i , θ i , φ i ) with intensity I i affects the internal volume of the cells ΔΩ (i) = [0 ≤ ρ ≤ ρ i + 3σ ρ , θ i ± 3σ θ , φ i ± 3σ φ For a cell M that is wholly or partially within ΔΩ, the conditional occupancy probability changes by ΔPpos(i)(M)=W(Ii)∫ΔΩ(i)∩Mp(i)(ρ,θ,φ)dΩ where W(I i ) is an empirical weighting factor that considers stronger detections as having a greater contribution to the occupancy probability compared to weaker detections.

[0032] To simplify storage and increase update speed, the occupancy grid stores the log(rates) rather than the rates. The log(rates) for a positive ISM storage bin can be retrieved using... Ln(M)=Ln−1(M)+∑ilog(ΔPpos(i)(M)PFA) will be updated.

[0033] The positive ISM can be calculated using brute force by calculating the integral: Ppos(i)(M)∝∫ΔΩ(i)∩Mppos(i)(ρ,θ,φ)dΩ which can prove to be inefficient and time-consuming. However, an approximation of the interval can be calculated by separating or factoring Eq. (8) into its angular and radial components, as in (9): ∫ΔΩ(i)∩Mp(i)(ρ,θ,φ)dΩ=∫ΔΩ(i)∩Mpang(i)(θ,φ)cos θ dθdφ×∫ΔΩ(i)∩Mprad(i)(ρ)ρ2dρ≈Pang(θi,φi,M)×Prad(ρi,M) shown

[0034] The angular and radial components can be considered independent of each other. Therefore, angular probability values ​​can be determined for the angular component, and radial probability values ​​can be determined from the radial component. These probability values ​​can then be multiplied to determine a probability value for a storage compartment of the positive ISM. Factoring Eq. (9) allows for the prior calculation of P. ang (θ i , φ i , M) and P rad (ρ i , M) values, so that these values ​​can be stored in a database and quickly retrieved. To calculate the probabilities, the values ​​are taken from the database and multiplied. In the radial direction, the behavior of +ΔLkn using the in the Fig. 3A and Fig. The probability diagrams shown in 3B postulate this.

[0035] Fig. Figure 3A (state of the art) shows a radial probability diagram 300 for creating or updating an occupancy grid based on the position of an object. For illustration, the object is located at a range or distance r from the sensor. The probability associated with detection at range r is given by a probability of 1. For distances between the sensor and the object, the probability value is zero, since detection at range r indicates no objects between the sensor and the object. The remaining distance between the object and the maximum range can be assigned a probability value of 1 / 2, indicating that there is an even probability that another object is located between the object and the maximum range, even if it remains undetected.

[0036] Fig. Figure 3B (state of the art) shows a log(ratio) model for a radial probability diagram 302 of an object located at range M0 from the sensor. The probability associated with detection at M0 is given by a positive probability |α|. Since detection occurs at M0, the probability that an object lies between the sensor and M0 is low. Otherwise, the object would be detected at a different radial position. Thus, a negative probability -|β| is assigned to the radial distances between the sensor and M0. A zero probability value can be assigned to the remaining distance between the location M0 and the maximum range.

[0037] Fig. Figure 4 shows an occupancy grid 400 illustrating co-detection and non-detection regions. The occupancy grid 400 shows a cross-section through an area of ​​the occupancy grid, highlighting areas of positive detection (co-detection regions) and negative detection (non-detection regions). The positive regions 406 and 408 are assigned probability values ​​or log(ratio) values ​​using rules from above with respect to the Fig. The positive ISM values ​​mentioned in points 2 and 3 are relevant here. Illustrative areas 410 and 412 are areas where there are no detections. Therefore, it is useful to be able to update the grid probability values ​​to represent missing detections in relation to the corresponding azimuth and elevation. The negative ISM discussed here updates the occupancy grid 400 for non-detection areas.

[0038] The negative ISM expresses the fact that if a particular direction in space contains no detections, the occupancy probability within the corresponding solid angle storage tray should decrease or fall. The negative ISM induces negative changes in the occupancy probability in cells of the occupancy grid that lack detections in the corresponding arrival direction (non-detection direction). Therefore, a probability value in a selected storage tray of the occupancy grid may be updated and reduced due to the presence of a detection in the direction corresponding to the storage tray during a previous time step if no detection occurs in that direction during the most recent time steps. The log(probabilities) for a direction with no detection can be calculated using Eq. (10): ΔLOneg(ρ)=log(1−pe(ρ)1−P0) for the: pe(ρ)=β(ρmax−ρ) is represented where ρmax where P0 is a maximum range of the sensor and P0 is a model parameter of the sensor. Since the derivation probability decreases with distance for any radar, the negative offset of Eq. (10) for storage compartments of the occupancy grid increases at greater ranges. In an alternative embodiment, the negative ISM can assume a linear decay of the log ratios with increasing range, as in Eq. (12): ΔLneg(ρ)=−|α|−|β|(ρmax−ρ) shown

[0039] Fig. Figure 5 shows an illustrative quad-tree structure 500 used to store and update probability values ​​for a negative ISM. The quad-tree structure comprises a plurality of levels, each level containing nodes in the form of solid angle storage compartments. On each level, the plurality of nodes for that level cover a selected solid angle of the sensor. In one embodiment, the selected solid angle is the sensor's field of view. A node or surface area on a level is linked to nodes on a lower level by a branching operation. With reference to level 1, area A covers a selected solid angle of the sensor. Area A can be subdivided by the branching operation into four separate solid angle areas (A1, A2, A3, A4) on the second level. The totality of the solid angle areas (A1, A2, A3, A4) on the second level encompasses area A on the first level.Each of these solid angle regions (A1, A2, A3, A4) can be further subdivided into four solid angle regions on a third level by the branching process. For illustration, solid angle A4 is divided into four solid angle regions (A1, A2, A3, A4, A4, A5, A6, A7, A8, A9, A1, A2, A3, A4) on the second level. 41 , A 42 , A 43 , A 44 ) subdivided. This process continues until a level is reached where no further subdivision of the solid angle is possible or desirable, either due to a resolution limit of the sensor or a resolution standard selected by a user.

[0040] Fig. Figure 6 illustrates the quad-tree structure of Fig. 5, to fill an occupancy grid 600 using a negative ISM. An illustrative solid angle 602 of an angular component of an occupancy grid 600 is shown on the left with two detections. On the right, the solid angle 602 of the angular component is shown. Storage compartments for which the detections exist are marked for illustration. A section of the solid angle 602 is covered by area A on a first level of a quad-tree structure. Using the branching operation, area A is subdivided into smaller areas A1, A2, A3, and A4.

[0041] The detection area is present in areas A2 and A3, while areas A1 and A4 have no existing detections. To create the negative ISM, negative ISM probabilities, as described in Eqs. (8)-(10), are entered into those storage bins (at the lowest level of the occupancy grid) that are within areas A1 and A4. However, areas A2 and A3 are subdivided at the next level of the quad-tree structure because detections are found in these areas.

[0042] To simplify the explanation, only area A2 will be discussed. Area A2 (a second-level node) is divided into areas A 21 , A 22 , A 23 and A 24 subdivided. The areas A 21 , A 22 and A 24 are free of detections. Therefore, suitable negative ISM probabilities can be entered into the corresponding storage areas. Area A23 However, it includes recognition and is therefore divided into areas on the next (third) level of the quad-tree structure.

[0043] The process continues as follows: If an area on one level of the quad-tree structure has no detections, a probability value for the negative ISM is assigned to the containers associated with the area; if an area has a detection, the area is subdivided into smaller areas on the next level of the quad-tree structure.

[0044] Area A 23 is therefore divided into areas A 231 , A 232 , A 233 and A 234 subdivided. For illustrative purposes, these areas are located on the bottom level of the quad-tree structure. Therefore, areas A 231 , A 232 and A 234 Negative ISM probability values ​​were assigned. Area A 233A value can be assigned with a positive ISM value, calculated using Eq. (9). This procedure can also be used to determine the area A3 in relation to the other in Fig. The 7 depicted recognitions can be subdivided.

[0045] In various embodiments, the processor 44 generates the negative ISM using the quad-tree structure of the Fig. 5 and Fig. 6. The negative ISM values ​​can be stored in a lookup table for better overview. Using the quad-tree structure increases the speed and efficiency of calculating the negative ISM.

[0046] After the negative ISM and the positive ISM have been created and / or updated, they are combined into a dynamic occupancy grid. The dynamic occupancy grid therefore contains bins to which a dynamic probability is assigned; that is, the probability associated with the bin changes with each new time step or each new incoming set of detections. For a bin in which a detection was recorded in an initial time step, but no detections were recorded in subsequent time steps, the current probability is modified by the negative ISM so that the bin has an associated probability that reflects the fact that no detections have been recorded more recently.In other words, the probability assigned to the storage compartment based on the detection in the first time step is reduced over time by corrections using the empty space weight.

[0047] Fig. Figure 7 shows a positive ISM 700, a negative ISM 702, and a dual ISM 704 for the occupancy grid, which combines the values ​​of the positive ISM 700 and the values ​​of the negative ISM 702. The negative ISM 702 affects all trays in the non-detection directions within the field of view, but not trays in co-detection directions. On the other hand, with the positive ISM 700, the log(rates) are zero, except within a narrow angle around a detection. Within this angle, the log(rates) are as shown by the radial diagram of Fig. 3b shown.

[0048] By updating the occupancy grid using the positive ISM 700 and the negative ISM 702, the probability values ​​assigned to the storage compartments of the occupancy grid change with each new time step or each new incoming detection set to reflect the current state of the vehicle's environment.

[0049] Fig. Figure 8 shows a flowchart 800 illustrating a method for generating an occupancy grid that dynamically represents the vehicle's environment 10. A detection system, such as a radar system 802, receives one or more detections 806. The radar system 802 includes a sensor that receives detections relating to objects within the vehicle's environment during a multitude of time-separated frames. The radar system 802 receives signals or detections 806 based on an object's range, azimuth, and elevation, as well as its Doppler frequency or velocity. In angular directions, the detections 806 are defined over a solid angle determined by the sensor's azimuth and elevation limits. An odometer 804 or other suitable speedometer provides a speed of the vehicle 10 to a self-motion estimation module 808.The vehicle's movement is provided to the radar detection system to separate the detections into a static set (810) and a dynamic set (812). The static set (810) includes objects within the environment that are not moving or are stationary, such as parked vehicles, buildings, signs, and other fixed structures. The dynamic detections (806) relate to objects that are moving within the environment, such as moving vehicles, pedestrians, cyclists, etc.

[0050] The static set 810 of detections and the dynamic set 812 of detections are provided to a clustering and outlier removal module 814, which filters the detections to remove noisy signals from both sets. Furthermore, the detections are grouped into clusters based on their relative proximity to one another. Outlier detections are generally not considered.

[0051] The clustering and outlier removal module 814 provides the filtered static set 816 of detections and the filtered dynamic set 818 of detections to an empty space update module 828. The empty space update module 828 creates an occupancy grid that records the locations where there are no detections within a selected frame. The empty space data sets, along with the updated static occupancy grid 826, can be used to update a dynamic occupancy grid 830 to reflect the current state of objects in the vehicle's environment in the form of a dynamic occupancy grid 832.

[0052] The filtered static detections 816 are provided to an update loop 820, which updates a static occupancy grid to reflect the changing nature of the environment. The update loop 820 includes a static occupancy grid history 822, which stores the previous version of the static occupancy grid. The static occupancy grid history 822 and the vehicle speed are used to resample the static occupancy grid. The resampled occupancy grid 824 and the filtered static detections 816 are used to generate an updated static occupancy grid 826. For a subsequent update step, the updated static occupancy grid 826 is used as the static occupancy grid history 822.The updated static occupancy grid 826 is provided to a dynamic occupancy update module 830 to provide a dynamic occupancy grid 832.

[0053] The Fig. 9 and Fig. Figure 10 illustrates differences in radar detection over time between a system that does not use the negative ISM for the occupancy grid and an occupancy grid disclosed herein that does use the negative ISM. Fig. Figure 9 shows a representation of an illustrative radar field in a scenario where an object passes in front of the vehicle. The radar field is formed from an occupancy grid that does not use the negative ISM. The object is located at a first location at time t0, a second location at time t1, and a third location at time t2. However, at time t2, the radar system shows a persistent stripe 902 of the object, extending from the object's location at time t0 to its location at time t2. Therefore, the persistent stripe 902 indicates the presence of an object at time t2 in locations where the object is no longer present.

[0054] Fig. Figure 10 shows the representation of another illustrative radar field in the same scenario as Fig. 9, wherein the radar field is formed from an occupancy grid that uses the negative ISM disclosed herein. Instead of a persistent strip 902, as in Fig. At time t0, the radar field shows an object at location 1002. At a later time t1, it becomes apparent that the objects have moved from location 1002 to the new location 1004. At a later time t2, the radar field shows that the object has moved again, this time from location 1004 to location 1006. Therefore, at each time step, the radar field only shows a localized representation of the object that is suitable for that particular time step.

Claims

[1] Method (800) for mapping an environment, comprising: Calculating a multitude of radial components and a multitude of angular components for a positive inverse sensor model (ISM) of an occupancy grid (200, 400, 600); Receiving a detection (204, 212) at a sensor (202) of an object in an environment surrounding a vehicle (10); Selecting a radial component corresponding to a detection area from the multitude of radial components and selecting an angular component corresponding to a detection angle from the multitude of angular components; Multiplying the selected radial component and the selected angular component to generate an occupancy grid (200, 400, 600) for detection; and Mapping the surroundings using the occupancy grid (200, 400, 600). [2] Method (800) according to claim 1, wherein the ISM further comprises a negative ISM (702) and a probability value of the negative ISM (702) is assigned using a quad-tree structure (500) of the occupancy grid (200, 400, 600). [3] Method (800) according to claim 2, wherein a node of the quad-tree structure (500) corresponds to a solid angle and a first level of the quad-tree structure (500) includes solid angle storage compartments (210, 218) which are combined to cover a field of view of the sensor (202). [4] Method (800) according to claim 3, further comprising dividing a solid angle storage compartment (210, 218) on a plane of the quad-tree structure (500) into a plurality of solid angles (602) on a plane of the quad-tree structure (500), if the solid angle storage compartment (210, 218) on the plane includes the recognition. [5] Method (800) according to claim 4, further comprising assigning a probability value to a solid angle storage compartment (210, 218) for which there is no recognition (204, 212), wherein the probability value reflects an absence of recognition in the solid angle storage compartment (210, 218). [6] System (100) for mapping the environment of a vehicle (10), comprising: a sensor (202) designed to detect an object in an environment surrounding the vehicle (10); and a processor (44) designed to: Calculating a multitude of radial components and a multitude of angular components for a positive inverse sensor model (ISM) of an occupancy grid (200, 400, 600); Selecting a radial component corresponding to a detection area (A) from the multitude of radial components and selecting an angular component corresponding to a detection angle from the multitude of angular components; Multiplying the selected radial component and the selected angular component to generate an occupancy grid (200, 400, 600) for detection (204, 212); and Mapping the surroundings using the occupancy grid (200, 400, 600). [7] System (100) according to claim 6, wherein the processor (44) is further configured to assign a probability value of a negative ISM of the occupancy grid (200, 400, 600) using a quad-tree structure (500) of the occupancy grid (200, 400, 600). [8] System (100) according to claim 7, wherein a node (A1, A2, A3) of the quad-tree structure (500) corresponds to a solid angle (602) and a first level of the quad-tree structure (500) includes solid angle storage compartments (210, 218) which together cover a field of view of the sensor (202). [9] System (100) according to claim 8, wherein the processor (44) is further configured to divide a solid angle storage compartment (210, 218) on a plane of the quad-tree structure (500) into a plurality of solid angles (602) on a plane of the quad-tree structure (500), when the solid angle storage compartment (210, 218) on the plane includes the recognition. [10] System (100) according to claim 9, wherein the processor (44) is further configured to assign a probability value to a solid angle storage compartment (210, 218) for which there is no recognition, wherein the probability value reflects an absence of recognition in the solid angle storage compartment (210, 218).

Images