Method and control arrangement for range estimation based on simulation

Abstract

The present disclosure relates to a computer-implemented method for determining a range ( d act ) of a sensor arrangement (11) arranged to monitor an environment of a vehicle (1). The computer-implemented method comprises simulating (S1) the environment during driving of the vehicle (1) using map data and / or pre-recorded terrain models together with shape models of perceived dynamic objects and estimating (S2), based on the simulating, a count of expected sensor detections, within one or more regions of the environment. The computer-implemented method further comprises obtaining (S3), from the sensor arrangement, observed sensor detections within the one or more regions (31) and determining (S4) the range by comparing the estimated count of expected sensor detections with a count of the observed sensor detections within the one or more regions. The disclosure also relates to a corresponding control arrangement (10) and computer program, and to a vehicle comprising the control arrangement.

Description

[0001]Method and control arrangement for range estimation based on simulation Technical field The present disclosure relates to a computer-implemented method for determining a range of a sensor arrangement arranged to monitor an environment of a vehicle. The disclosure also relates to a corresponding control arrangement and computer program, and to a vehicle comprising the control arrangement. Background An autonomous vehicle relies on sensors on the vehicle to navigate safely. During autonomous driving it is essential to know the effective range of the sensors at all times. Dirt, precipitation, air moisture, physical occlusions (e.g., buildings obstructing the sensors), and light conditions constantly affect the performance of the sensors. Using static assumptions on the effective range of sensors may lead to an overestimated ability to detect obstacles during situations with degraded performance. This may lead to accidents because an autonomous vehicle does not realize the risk that may lurk in the areas beyond the effective range. One possibility is to estimate sensor range by detecting landmarks and comparing distances between neighboring landmarks to the known distances (i.e., from map data) between these landmarks (see DE102018003784A1). However, such methods typically require long-term accumulation and filtering of data related to the detected landmarks and are therefore time and resource consuming. The filtering also makes sudden degradation effects that take long to discover. Hence, there is a need for improved methods for sensor range estimation that can be used in autonomous driving. It is an objective of the present disclosure to provide a method for determining range of sensors arranged to monitor an environment of a vehicle. It is also an objective to provide a method that can separate physical occlusions from degraded sensor performance. These objectives and others are at least partly achieved by the computer-implemented method, control arrangement, and vehicle according to the independent claims, and by the embodiments according to the dependent claims. According to a first aspect, the disclosure relates to a computer-implemented method for determining a range of a sensor arrangement arranged to monitor an environment of a vehicle. The computer-implemented method comprises simulating the environment during driving of the vehicle using map data and / or pre-recorded terrain models together with shape models of perceived dynamic objects and estimating, based on the simulating, a count of expected sensor detections, within one or more regions of the environment. The computer-implemented method further comprises obtaining, from the sensor arrangement, observed sensor detections within the one or more regions and determining the range by comparing the estimated count of expected sensor detections with a count of the observed sensor detections within the one or more regions. The computer-implemented method will grant nearly immediate results for estimating the effective range as sensor detections can be compared instantaneously. In particular, compared to other methods that typically include long-term accumulation and statistics gathering about detected objects or landmarks, the computer-implemented method will yield near immediate results. In some embodiments, the one or more regions represent different distance ranges from the sensor arrangement. By analysing the number of hits within a region corresponding to a certain distance range, it is possible to determine whether the distance range is within or outside coverage of the sensor arrangement. In some embodiments, a sensor detection is defined as a lidar and / or radar hit within the one or more regions. In some embodiments, a sensor detection is defined as a pixel, or group of pixels, in an image captured by an image sensor, depicting a surface or object located within the one or more regions. In some embodiments, a sensor detection is defined as any other signal reflected from a surface or object in the one or more regions, wherein the signal is detectable by the sensor arrangement. Hence, a sensor detection could be any metric indicating presence of solid or liquid material at a predefined point, or area, in space. For example, objects, water drops, or dust may be detected. In some embodiments, the estimating comprises excluding sensor detections in occluded regions that are occluded from a field of view of the sensor arrangement. Hence, by using this method in conjunction with occlusion predictions it is possible to handle the risks with degraded sensor performance in a safe way. In some embodiments, the determining comprises comparing the observed and expected sensor detections using histograms, wherein the observed sensor detections are binned in the histogram based on range from the sensor arrangement. Using a histogram is an efficient way to implement the computer- implemented method. In some embodiments, the determining comprises establishing that a region is out of range upon on a difference between estimated and observed sensor detections within that region falling below a tolerance threshold. Hence, different thresholds may be defined depending on for example the expected sensor detection rate. In some embodiments, the simulating comprises creating an onboard 3D-model of the environment. Various 3D-simulations can be used as a basis for the estimation. In some embodiments, the expected sensor detections comprise detections of road surface, static landmarks and / or traffic participants included in the simulating. The computer-implemented method may utilise detections of various objects visible in map data and / or pre-recorded terrain models. In some embodiments, the one or more regions cover a road the vehicle is travelling. In some embodiments, the one or more regions cover an area adjacent to, or around, the road the vehicle is travelling. Also simulating the terrain makes it possible to exclude the lack of detections in areas that are physically occluded from the sensors. In certain situations, with very few landmarks or with hilly / winding roads, it is possible to separate the physical occlusions from a degraded sensor. This means it is possible to detect degraded performance in e.g., sudden rain / snow / hail or water splash blocking a sensor, dirt clouds, sudden interference from a light or radio source etc. In some embodiments, the computer-implemented method comprises operating the vehicle based on the simulating, wherein a region outside the determined range is treated as occluded from the one or more sensors. By using the computer- implemented method, a correct range of the sensor arrangement can be used when operating the vehicle whereby risks associated with degraded sensor performance can be handled in a safe and efficient way. According to a second aspect, the disclosure relates to a computer program comprising instructions which, when the computer program is executed by a computer, cause the computer to carry out the computer-implemented method according to the first aspect. According to a third aspect, the disclosure relates to a computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the computer-implemented method according to the first aspect. According to a fourth aspect, the disclosure relates to a control arrangement configured to perform the computer-implemented method according to the first aspect. According to a fifth aspect, the disclosure relates to a vehicle comprising the control arrangement according to the fourth aspect. Corresponding effects as for the first aspect can be achieved by the second to fifth aspects. Brief description of the drawings The embodiments disclosed herein are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings. Like reference numerals refer to corresponding parts throughout the drawings, in which: Figs.1A-C illustrate the difference between real and expected sensor range. Fig.2A illustrates an ego vehicle in an environment with other traffic participants. Fig.2B illustrates an ego vehicle in an environment with other traffic participants and various occluding objects. Fig. 3A illustrates an ego vehicle where the proposed technique may be implemented. Fig.3B illustrates a control arrangement of the vehicle in Fig.3A in further detail. Fig.3C illustrates a sensor of the vehicle in Fig.3A in further detail. Fig. 4 is a flow chart of a computer-implemented method for operating a vehicle according to the first aspect. Figs.5A-B illustrate histograms showing the number of detections per range. Detailed description The present disclosure describes methods for estimating the range of a sensor arrangement comprising one or more sensors. Figs. 1A-C illustrate how the environment can influence the effective sensor range ^^. Fig. 1A illustrates the expected sensor range ^^^௫^under normal environmental conditions. Fig. 1B illustrates how the real sensor range ^^^^௧is affected by clouds and moisture. Fig. 1C illustrates how an object (here a sign), which would normally be detected is occluded because of the actual sensor range ^^^^௧being limited. The occluded region 3 caused by the clouds and moisture shall typically be considered occluded by a motion planner of the vehicle 3. It is herein proposed to estimate the range of sensors in a vehicle not only based on detection of the road surface and / or static landmarks from the vehicle sensors, but to also use information from an onboard sensor simulation of the environment. More specifically, a simulation of the environment is used to predict what the sensors will see. By comparing real sensor readings and the predictions based on the simulation, it is possible to determine if landmarks / road surface that one would expect to see based on the simulated model, are detected in real life. This can be accomplished by comparing observed sensor detections of the sensors with predicted sensor detections. In addition, physical occlusion can be generated for example by terrain and buildings, as well as from other road users. It is important to separate the physical occlusions from the degraded performance of the sensor due to the factors mentioned above. Hence, it is also proposed that knowledge about occluded objects, which is available from the simulation, is also considered when determining the range. The proposed technique is based on simulating the environment and prediction of the traffic in the environment. For better understanding of the proposed technique, an example model for determining a trajectory of a vehicle based on predicted traffic in the environment will first be described with reference to a general example. FIG. 2A illustrates an environment of an ego vehicle 1 in an upcoming T- intersection. Traffic participant 2 is shown driving on lane 4a of lanes 4a, 4b in the direction of traffic (illustrated with arrows). Traffic participant 2 does not occlude any relevant part of the road for the ego vehicle 1. The traffic participant 2 is visible in the field of view of sensors 11 (Fig. 3A) on the ego vehicle 1 and the traffic can easily be predicted. Once traffic is predicted, a feasible and collision-free trajectory 8 can be calculated for the ego vehicle 1. Simulating the environment The environment comprises a set of traffic participants, including an ego vehicle 1 and other participants 2. The simulation involves using reachable sets as predictions of other traffic participants (including all types of traffic participants such as trucks, cars, bikes etc.). In other words, a traffic participant 2 at a time ^^ is represented as a set of states ^^^^^^ ⊂ ℝ^^. Each traffic participant 2 is modeled as^^ ≔ ^^^ெ,^^ெ,^^ெ ^, where ^^^^^^^ ൌ ^^ெ൫^^^^^^,^^^^^^൯ denotes the state dynamics,^^ெ ⊆ ℝ^^ denotes the admissible set of states, and ^^ெ ⊆ ℝ^^ the admissible set ofinputs. An occupied space ^^^^^^is a set of points in the ^^^^-plane occupied by othertraffic participants, namely ^^^^^^ ൌ ^^^^^^^^ ை௫௬^^^ ^^^^^, where ^^ை^^^^ is the set of all othertraffic participants’ states at time ^^. The field of view ℱ^^^^ is the set of points on the^^^^-plane, ℱ^^^^ ⊂ ℝଶ, within direct line of sight of a vehicles’ sensors, and within thesensors’ range. The notion of occupied space and field of view are given from the ego vehicle’s perspective. The set of valid states for a traffic participant is ^^௩^^^ௗൌ^^^ ∈ ∣ ^^^^^^^^௫௬^^^^ ∈ ℒ ⋀ ^^ ∈ ^^^, where ℒ ⊆ ℝଶ is the legal drivable area and ^^ ⊆ℝ^are constraints (e.g., a heading constraint or a velocity limit ^^௩). In one embodiment, it is assumed that traffic participants 2 such as cars, trucks and motorcycles can be modelled accurately using a non-linear bicycle model (“bic”) described in R. Rajamani, Vehicle Dynamics and Control, Mechanical Engineering Series, Boston, MA, Springer, US, 2012, ^^^^^≔ ^^^ெ,^^ெ,^^ெ^, where The model has state variables ^^, ^^, an orientation ^^, and a velocity ^^. The length between the front and rear axle, the wheelbase, is denoted ^^. As input, the modeltakes a steering angle ^^ and an acceleration ^^. ^^^,^^^, ^^^,^^௫, and ^^ఏ are constraintson the acceleration ^^ and steering angle ^^. The x-coordinate may be assumed to be aligned with driving direction of the ego vehicle’s lane, and the heading of allother traffic participants are assumed to be in the interval ^ ^^⁄ 2during the relevant time horizon. The reachable set of states of the non-linear bicycle model may be difficult to compute or directly overapproximate. However, reachable sets can be computed for several linear model abstractions and then combined. The model abstractions are a type of model of how the states of traffic participants change over time. They are simplifications of the original model (equation 1). It might also be desired to overapproximate the set of reachable states to ensure safe properties of the system. The reachable set of states may be over-approximated by adding the system’s (equation 1) homogenous and inhomogeneous solutions together and considering infinitely many terms in the sum. In one example embodiment, a first model abstraction ^^௩^^, that over-approximates the reachable set of states in (^^,^^), is combined with a second model abstraction, ^^^^^, that over-approximate thereachable set of states in (^^, ^^).The first model abstraction, ^^௩^^ ∶ൌ ^^^௩^^ , ^^௩^^ , ^^௩^^ ^ is a simple integrator model.The model is linear and only has two states, ^^ and ^^, and two inputs ^^௫and ^^௬. Since the states are directly controlled by the input, the system matrix is ^^௩^^ൌ ^^. With ^^௩^^possible velocities are used to overestimate which positions a traffic participant can reach (in ^^^^-coordinates). Hence, if a traffic participant has anyposition (^^^,^^^) in the set ^^^ at a first time instance then its position (^^ଶ, ^^ଶ) attime ^^ଶmust be in the set ^^ଶ. Then an over-approximation of ^^ଶ, denoted calculated.A second model abstraction, ^^ ∶ൌ ^^^^^^,^^^^^,^^^^^^^^ ^ describes trafficparticipants’ dynamic behavior along the x-axis as a double integrator system. It has two states, ^^ and ^^, and one inputs ^^௫. Being linear and time-invariant, themodel may be represented in the state-space form ^^^ ^^^^ ൌ ^^^^^^^^^^ ^ ^^^^^^, with^^^^^ ൌ ^0 10 0^.Since the bicycle model’s reachable states in (^^, ^^) are governed by the doubleintegrator dynamics, the abstraction is to change of input from ^^ to ^^௫. To translate the constraints on heading and velocity, it is identified that maximum reach in ^^ is achieved with maximal acceleration and zero heading, and minimum reach is achieved with minimum acceleration and maximal heading. With ^^^^^possible accelerations are used to overestimate which velocities a traffic participant can reach at different positions (in ^^^^-coordinates). Hence, if a traffic participant hasany position and velocity combination (^^^, ^^^) in the set ^^ ^ at a first time instance^^^ then its position and velocity (^^ଶ, ^^ଶ) at a second time instance ^^ଶ must be in theset ^^ଶ. Then an over-approximation of ^^ଶ, denoted ^^^^^^^^^, ^^^^^^^^௫௩^^^^^^, ^^ଶ – ^^^^), iscalculated. In some embodiments, an over-approximation of the reachable set of states from ^^௩^^is the initial set of states, Minkowski summed with the time-scaled input set. In some embodiments, an over-approximation of the reachable set of states from ^^^^^is a skewing of the initial set of states in ^^ and Minkowski summing it with a time- scaled version of the input set. With these insights, the first and second abstractions can be combined to over-approximate the original bicycle model. Thereby, thereachable set of states at a time ^^ ^ ^^^^, for a traffic participant currently in the setof states ^^ఛ, can be over-approximated such that: where the mapping ^^௫௬௩ ∶ ^^^, ^^, ^^^ ^ ^^^, ^^, ^^^ to correct the element order after theCartesian products with the missing dimensions. The over-approximated reachableset of states is denoted ℛ^^^^^^^ ,^^ఛ,∆^^^. Hence, for each traffic participant an over-approximated reachable set of states can be determined, that can be used when planning a trajectory for the ego vehicle 1. FIG.2B illustrates the same environment of ego vehicle 1 as in Fig.2A, but with occluding objects in the field of view of the sensors 11 on ego vehicle 1. The occluded objects here are both other traffic participants 2 and a stationary object 6. The occluding objects occlude parts of the environment, referred to as occluded regions 3 (i.e. an occluded area), from the field of view of the sensors 11 on the ego vehicle 1. Ego vehicle 1 needs to predict possible traffic participants also in these occluded regions 3 to make a feasible and safe trajectory. Hence, the possible traffic participants also need to be included in the model. This is done by creating one or more “ghost” traffic participants in the occluded regions 3, referred to as “possible traffic participants”. Fig. 3A illustrates an ego vehicle 1, here a truck, where the proposed technique may be implemented. The ego vehicle 1, herein referred to as simply the vehicle 1, may be fully autonomous or at least partly manual. The vehicle 1 comprises a plurality of electric systems and subsystems. For simplicity only some parts of the vehicle 1 that are associated with the proposed method are shown in Fig.3A. Thus, the illustrated vehicle of Fig.3A comprises a control arrangement 10, a plurality of sensors 11, a propulsion system 12, a braking system 13 and an autonomous driving function 14. The plurality of sensors comprises one or more sensors 11 configured to sense the environment of the vehicle 1. The one or more sensors 11 may be embodied as sensor devices. For example, the one or more sensors 11 may include Lidar, Radar, image and proximity sensors. The one or more sensors 11 may be configured to detect objects in the environment, such as other traffic participants 2 and other objects 6. The one or more sensors 11 are also configured to estimate size, position and velocity of detected objects using commonly known techniques. Figure 3C illustrates a sensor 11 in more detail. The sensor 11 typically has a field of view, FoV, ^^. The FoV refers to the extent of the observable area that a sensor, camera, or other optical device can capture at any given moment. A certain FoV can be expected based on the sensor hardware, i.e., sensor type. Hence, a range of a sensor 11 is herein defined as a distance within which accurate sensor readings are expected. In other words, the term sensor range ^^ refers to the maximum distance or area over which a sensor can effectively detect, measure, or monitor a particular parameter or stimulus. Various factors may influence the sensor range, such as the environment in which the sensor 11 is operating. For example, the range may be affected by humidity, light, dirt on the sensor 11 etc. Hence, one or more regions 31 within the sensor range may be occluded. In other words, the actual sensor range ^^^^௧may be shorter than an expected sensor range ^^^௫^. The dashed lines illustrate examples of ^^^^௧. The propulsion system 12 is designed to propel the vehicle. The propulsion system typically comprises a combustion engine and / or an electrical motor. The braking system 13 is arranged to decelerate vehicle 1. The braking system typically comprises several different braking systems that interact, such as main brakes and auxiliary brakes. The control arrangement 10 is configured to operate a first vehicle 1. The control arrangement comprises an autonomous driving function 14. The autonomous driving function 14 is configured to control driving operation of the vehicle 1. In some embodiments the vehicle 1 is fully autonomous. In these embodiments the autonomous driving function 14 may comprise a mission handler configured to receive missions from an operator or off board control system. In these embodiments, the autonomous driving function 14 is configured to calculate a trajectory for the vehicle to drive and to control driving along the trajectory. The autonomous driving function 14 controls operation of the vehicle by sending commands to the propulsion system 12, to the braking system 13 and to other systems, such as to a steering system (not shown). In other embodiments, the vehicle 1 is at least partly manual. The autonomous driving function 14 may then be configured to control the vehicle 1 in certain situations, such as when the function is activated by a driver or as a safety system in order to avoid accidents. In some embodiments, the autonomous driving function 14 is an Advanced Driver-Assistance System, ADAS. Fig. 4 is a flow chart of computer-implemented method estimating a range of a sensor arrangement according to the first aspect. A sensor arrangement herein refers to either one individual sensor 11 or a combination of a plurality of sensors 11 (Fig. 3A). The computer-implemented method may be implemented as a computer program comprising instructions which, when the program is executed by a computer (e.g., a processor in the control arrangement 10 (Fig.3B)), causes the computer to carry out the computer-implemented method. According to some embodiments the computer program is stored in a computer-readable medium (e.g., a memory or a compact disc) that comprises instructions which, when executed by a computer, cause the computer to carry out the method. In some embodiments, the method is implemented in a control arrangement 10 of a vehicle 1, e.g., in the vehicle 1 of Fig.3A. The method can be performed continuously during autonomous driving of the vehicle 1. The method may be performed in an ongoing manner, such as repetitively or continually. The method may be performed for a plurality of individual sensors or sensor arrangements 11. Thereby, an accurate estimation of the sensor range can be used by the ADAS. In order to autonomously operate a vehicle 1, the environment of the vehicle 1 must be simulated to find a trajectory and avoid collisions. In other words, during operation a model of the physical environment is implemented, i.e. the environment is simulated. The model is continuously updated based on sensor data provided by the sensors 11. In other words, the method comprises simulating S1 the environment during driving of the vehicle using map data and / or pre-recorded terrain models together with shape models of perceived dynamic objects. Perceived dynamic objects corresponds to any object that may be simulated, such as traffic participants including vehicles, bicycles, pedestrians etc. Map data may herein refer to any prerecorded graphic representation of static features of the environment. Map data in autonomous driving typically refers to highly detailed digital maps that provide critical information about road networks, including lane configurations, traffic signs, and road geometry. Official authorities may provide detailed maps and surveys that include information on road layouts, traffic signs, and other relevant data. Pre-recorded terrain models are detailed digital representations of the Earth's surface that capture topographical features, including elevation, slope, and landforms. These models are created using data collected from various sources such as LiDAR, aerial imagery, and satellite data, and are used in applications like autonomous driving to enhance navigation, improve localization, and provide a better understanding of the driving environment by integrating these models with real-time sensor data. In addition data captured by the sensors 11 are used in the simulating to create shape models of perceived dynamic object. For example, the sensor data is used to model other traffic participants. For example the simulating S1 comprises creating a model as described above. In other words, in some embodiments, simulating S1 comprises creating an onboard 2D or 3D-model of the environment. Because the digital implementation of the mathematical model provides a representation of the physical environment, the model may be used to predict what the sensors 11 will see. For example, if there is an object or surface in a certain region and a lidar signal is directed towards that region, then one may beforehand calculate how many reflections the object or surface would generate. Thus, the proposed technique is based on the insight that a count of sensor detections in individual regions may be estimated using the model. Hence, compared to known solutions that utilizes only information about objects known from, for example maps, the proposed technique also utilizes information about what the sensors have seen before. Also, when determining range of one sensor, then information from other sensors can be used. In other words, all information available through the simulation can be used. In other words, the method comprises estimating S2, based on the simulating S1, a count of expected sensor detections, within one or more regions of the environment. A count herein refers to a number, i.e., a number obtained by counting how many sensor detections are received from each region of the one or more regions. The count may be a count for an individual sensor, such as a radar or lidar, or a total number of detections for several sensors. The estimating S2 is typically performed in several different regions. In some embodiments, the one or more regions 31 represent different distance ranges from the sensor arrangement, see Fig. 3C. The regions may correspond to the entire field of view of the sensor arrangement, as in Fig.3C. Alternatively, the regions may also correspond to different viewing angels. In some embodiments, the one or more regions cover a road the vehicle 1 is travelling. However, it is also possible that the one or more regions cover an area adjacent to or around the road the vehicle 1 is travelling. Depending on where the regions are located the expected sensor detections may originate from different types of objects. Hence, in some embodiments, the expected sensor detections comprise detections of road surface, static landmarks and / or traffic participants included in the simulating S1. A sensor detection herein refers to detection of solid material at a certain point. In some embodiments a sensor detection is defined as a lidar and / or radar hit within the one or more regions. In other words, a sensor detection can be a reflection of a signal, i.e., by an object within the one or more regions. For some sensors, like LIDAR and RADAR, the sensor detection is a reflection of a signal transmitted by the sensor itself. Other sensors rather detect reflections of ambient light (e.g., from the sun). In these embodiments, sensor detections could be pixels picturing or revealing the object. Hence, in some embodiments a sensor detection is defined as a pixel, or group of pixels, in an image captured by an image sensor, depicting a surface or object located within the one or more regions. Depending on the type of sensor arrangement a sensor detection could be defined as any signal reflected from a surface, or object, in the one or more regions and detectable by the sensor arrangement. When the count of sensor detections is estimated, one may also utilize other information, such as information about occluding regions 3 (Fig. 2B). If a certain region is occluded from the sensor arrangement, it means that there may not be any sensor detections corresponding to an object located within the occluded region, even if the object is within the range of the sensors. Thus, presence of occluded regions should be taken into account when estimating the expected number of detections. In other words, detections within occluded regions should be excluded. In other words, in some embodiments, estimating S2 a count of expected sensor detections, comprises excluding sensor detections in occluded regions that are occluded from a field of view of the sensor arrangement. The expected count of sensor detections shall then be compared to real sensor observations. In other words, the method comprises obtaining S3, from the sensor arrangement, observed sensor detections within the one or more regions. More specifically, the obtaining S3 involves investigating how many reflections are originating from the respective regions of the one or more regions. For example, if there is an object in a region, there would be many sensor detections. Especially, it is usually expected to get many detections of road surface. One can find the maximum range for which we get road surface detections and estimate the sensor range. However, this information is however not sufficient to draw conclusions about the full efficient range of the sensors. By comparing the estimated effective range with the model-based expected effective range it is possible to draw more detailed conclusions. Thus, the range can be more precisely determined by evaluating whether the sensor arrangement detects objects that it is expected to detect based on the simulation of the environment. In other words, the method comprises determining S4 the range by comparing the estimated count of expected sensor detections with a count of the observed sensor detections within the one or more regions. In principle, it means that if the number of real sensor detections is significantly lower than the expected value from the simulation, then the range is shorted than expected. In other words, the determining S4 comprises establishing that a region is out of range upon on a difference between estimated and observed sensor detections within that region exceeding a tolerance threshold. There are different ways of implemented the determining. One feasible way is to use a histogram. A histogram is a graphical representation of the distribution of numerical data. It is a type of bar chart that displays the frequency or count of data points falling within specified intervals, known as bins. In other words, in some embodiments, determining S4 comprises comparing the observed and expected sensor detections using histograms, where the observed sensor detections are binned based on range from the sensor arrangement. In other words, a histogram is created where each bin corresponds to a certain range (typically an interval) from the sensor arrangement. Figs.5A-B illustrate two different scenarios where histograms are used to show the number of detections per range. In these examples, the ratio between the simulated (curve 51) and actual (bars 52) number of road / landmark detections is compared. By putting a threshold on the ratio between the actual and expected number of detections it is possible to find a range at which the number of detections gets fewer than expected, thereby getting an estimate of the actual effective sensor range. In Fig.5A the vehicle 1 is driving on a straight road and the sensors are not occluded by any objects. The number of expected sensor detections in each region (curve 51) can then be estimated based on the simulation. The actual sensor detections are then binned based on range. For each range interval (e.g., one region 31 in Fig. 3C) the count of sensor detections is represented by a bar 52, where the height of the bar 52 corresponds to the number of sensor detections. Any deviation between the bars 52 and the curve 51 indicate that the corresponding distance is out of range. In Fig. 5A the bars 52 representing the three longest distances from the sensor arrangement are significantly lower than expected. Hence, it can be concluded that the actual sensor range ^^^^௧is reduced. In other words, the area between ^^^^௧and ^^^௫^can not be expected to be seen by the sensor arrangement. Hence, in this example, the actual range ^^^^௧of the sensors is shorter than the expected sensor range ^^^௫^(e.g., an expected range of the sensor arrangement calculated based on hardware limitations of the sensors). In the example of Fig. 5B the sensor arrangement is occluded by another traffic participant 2. This means that the expected number of sensor detections behind the vehicle is zero, as explained above. Hence, no conclusions about the range can be made. Instead, one has to wait until the traffic participant 2 has passed. Then the range can be estimated in the same way as in the previous example. In some embodiments, the method comprises operating S5 the vehicle based on the simulating S1, wherein a region outside the determined range is treated as occluded from the one or more sensors. The online estimated effective range of sensors can be used to reason about potentially unseen road users beyond the sensor range. Fig.3B illustrates a control arrangement 10 configured to operate a first vehicle 1. The control arrangement 10 may be arranged in the vehicle 1 (Fig.1). Control arrangement 10 comprises control circuitry to perform the method according to any one of the steps, examples or embodiments as described herein. The control arrangement 10 may include one or more Electronic Control Units (ECUs) connected to a controller area network (CAN). For example, the control arrangement 10 may be an Electrical Control Unit, ECU, of the ACC. More in detail, the control arrangement 10 comprises one, or more, computer(s) 101 and memory 102. The computer 101 comprises any hardware or hardware / firmware device implemented using processing circuity such as, but not limited to, a processor, Central Processing Unit (CPU), a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a System-on-Chip (SoC), a programmable logic unit, a microprocessor, an application-specific integrated circuit, or any other device capable of electronically performing operations in a defined manner. In some embodiments, the computer-readable medium may be a non-transitory computer- readable medium, such as a tangible electronic, magnetic, optical, infrared, electromagnetic, and / or semiconductor system, apparatus, and / or device. The computer-readable memory is for example one or more of the memories in the control arrangement 10. Hence, the proposed method may be implemented as a computer program. The computer program then comprises instructions which, when the computer program is executed by a computer, cause the computer to carry out the method according to any one of the aspects, embodiments or examples as described herein. In some embodiments the control arrangement 10 comprises a communication interface 103 configured to enable wireless communication with off-board devices, such as with other vehicles, road objects or with a data storage, such as a cloud server. The wireless communication may be performed using any suitable protocol for V2X communication. This communication may be performed via a Controller Area Network, CAN, or directly via an embedded modem. More specifically, the control arrangement 10 is configured to model the environment during driving of the vehicle using map data and / or pre-recorded terrain models together with shape models of perceived dynamic objects and to estimate, based on the simulating, a count of expected sensor detections, within one or more regions of the environment. The control arrangement 10 is further configured to obtain from the sensor arrangement 2, observed sensor detections within the one or more regions and to determine the range by comparing the estimated count of expected sensor detections with a count of the observed sensor detections within the one or more regions. The terminology used in the description of the embodiments as illustrated in the accompanying drawings is not intended to be limiting of the described method, control arrangement or computer program. Various changes, substitutions and / or alterations may be made, without departing from disclosure embodiments as defined by the appended claims. The term “or” as used herein, is to be interpreted as a mathematical OR, i.e., as an inclusive disjunction; not as a mathematical exclusive OR (XOR), unless expressly stated otherwise. In addition, the singular forms "a", "an" and "the" are to be interpreted as “at least one”, thus also possibly comprising a plurality of entities of the same kind, unless expressly stated otherwise. It will be further understood that the terms "includes", "comprises", "including" and / or "comprising", specifies the presence of stated features, actions, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, actions, integers, steps, operations, elements, components, and / or groups thereof. A single unit such as e.g. a processor may fulfil the functions of several items recited in the claims. The present disclosure is not limited to the above-described preferred embodiments. Various alternatives, modifications and equivalents may be used. Therefore, the above embodiments should not be taken as limiting the scope of the disclosure, which is defined by the appending claims.

Claims

Claims 1. A computer-implemented method for determining a range of a sensor arrangement arranged to monitor an environment of a vehicle (1), the computer-implemented method comprising: - simulating (S1) the environment during driving of the vehicle using map data and / or pre-recorded terrain models together with shape models of perceived dynamic objects, - estimating (S2), based on the simulating (S1), a count of expected sensor detections, within one or more regions of the environment, - obtaining (S3), from the sensor arrangement, observed sensor detections within the one or more regions, - determining (S4) the range by comparing the estimated count of expected sensor detections with a count of the observed sensor detections within the one or more regions.

2. The computer-implemented method according to claim 1, wherein the one or more regions represent different distance ranges from the sensor arrangement.

3. The computer-implemented method according to claim 1 or 2, wherein a sensor detection is defined as: - a lidar and / or radar hit within the one or more regions, - a pixel, or group of pixels, in an image captured by an image sensor, depicting a surface or object located within the one or more regions, or - any other signal reflected from a surface, or object, in the one or more regions and detectable by the sensor arrangement.

4. The computer-implemented method according to any one of the preceding claims, wherein the estimating (S2) a count of expected sensor detections, comprises excluding sensor detections in occluded regions that are occluded from a field of view of the sensor arrangement (2).

5. The computer-implemented method according to any one of the preceding claims, wherein the determining (S4) comprises comparing the observed and expected sensor detections using histograms, where the observed sensor detections are binned based on range from the sensor arrangement.

6. The computer-implemented method according to any one of the preceding claims, wherein the determining (S4) comprises establishing that a region is out of range upon on a difference between estimated and observed sensor detections within that region exceeding a tolerance threshold.

7. The computer-implemented method according to any one of the preceding claims, wherein the simulating (S1) comprises creating an onboard 3D- model of the environment.

8. The computer-implemented method according to any one of the preceding claims, wherein the expected sensor detections comprise detections of road surface, static landmarks and / or traffic participants included in the simulating (S1).

9. The computer-implemented method according to any one of the preceding claims, wherein the one or more regions cover a road the vehicle (1) is travelling.

10. The computer-implemented method according to any one of the preceding claims, wherein the one or more regions cover an area adjacent to or around the road the vehicle (1) is travelling.

11. The computer-implemented method according to any one of the preceding claims, comprising operating the vehicle based on the simulating (S1), wherein a region outside the determined range is treated as occluded from the one or more sensors.

12. A computer program comprising instructions which, when the computer program is executed by a computer, cause the computer to carry out the computer-implemented method according to any one of the preceding claims.

13. A computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the computer-implemented method according to any one of the claims 1 to 11.

14. A control arrangement (10) configured to estimate a range of a sensor arrangement arranged to monitor an environment of a vehicle (1), wherein the control arrangement is configured to: - model the environment during driving of the vehicle using map data and / or pre-recorded terrain models together with shape models of perceived dynamic objects, - estimate, based on the simulating (S1), a count of expected sensor detections, within one or more regions of the environment, - obtain, from the sensor arrangement, observed sensor detections within the one or more regions, - determine the range by comparing the estimated count of expected sensor detections with a count of the observed sensor detections within the one or more regions.

15. The control arrangement (10) according to claim 13 wherein the control arrangement is configured to perform the computer-implemented method according to any one of claims 2-11.

16. A vehicle (1) comprising the control arrangement (10) according to claim 14 or 15.

Images