WO2024115439A1 - Procédé de fonctionnement d'un système lidar, et système lidar - Google Patents

Procédé de fonctionnement d'un système lidar, et système lidar Download PDF

Info

Publication number
WO2024115439A1
WO2024115439A1 PCT/EP2023/083270 EP2023083270W WO2024115439A1 WO 2024115439 A1 WO2024115439 A1 WO 2024115439A1 EP 2023083270 W EP2023083270 W EP 2023083270W WO 2024115439 A1 WO2024115439 A1 WO 2024115439A1
Authority
WO
WIPO (PCT)
Prior art keywords
received signal
sampling
lidar system
sequence
threshold
Prior art date
Application number
PCT/EP2023/083270
Other languages
German (de)
English (en)
Inventor
Marco Heinen
Patrick REICHEL
Kai Fischer
Original Assignee
Valeo Schalter Und Sensoren Gmbh
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Valeo Schalter Und Sensoren Gmbh filed Critical Valeo Schalter Und Sensoren Gmbh
Publication of WO2024115439A1 publication Critical patent/WO2024115439A1/fr

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
    • G01S7/483—Details of pulse systems
    • G01S7/486—Receivers
    • G01S7/4865—Time delay measurement, e.g. time-of-flight measurement, time of arrival measurement or determining the exact position of a peak
    • G01S7/4866—Time delay measurement, e.g. time-of-flight measurement, time of arrival measurement or determining the exact position of a peak by fitting a model or function to the received signal
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/88—Lidar systems specially adapted for specific applications
    • G01S17/93—Lidar systems specially adapted for specific applications for anti-collision purposes
    • G01S17/931—Lidar systems specially adapted for specific applications for anti-collision purposes of land vehicles

Definitions

  • the application relates to a method for operating a lidar system, a lidar system, a vehicle with the lidar system and a use of the lidar system in a vehicle.
  • Modern vehicles have a large number of sensors whose data is used to provide driver information and/or driver assistance systems.
  • the sensors record the vehicle's surroundings and other road users. Based on the recorded data, a model of the vehicle's environment can be created and changes in this vehicle environment can be responded to.
  • a Lidar system has an optical transmitter and an optical receiver.
  • the transmitter can emit an optical transmission signal, which can be pulsed.
  • laser beams in the ultraviolet, visual or infrared range can be used in particular.
  • the receiver can receive the optical signal as an optical reception signal after reflection from an object in a detection area in the vicinity of the Lidar system.
  • the reception signal can be evaluated using the transmission signal by a computing unit of the Lidar system using a time-of-flight method and the spatial position and distance of the objects on which the reflection occurred can be determined. It is also possible to determine a relative speed.
  • Time-of-Flight (TOF) systems can be used in particular to determine the distance to objects.
  • Lidar systems are constantly being developed for various functions, e.g. for the acquisition of environmental information in the near and far range of vehicles, such as passenger cars or commercial vehicles.
  • Lidar systems can also serve as sensor systems for driver assistance systems, in particular assistance systems for autonomous or semi-autonomous vehicle control. They can be used in particular to detect obstacles and/or other road users in the front, rear or blind spot area of a vehicle.
  • the lidar system can be designed as a system that works with flashes of light, a so-called flash lidar.
  • An area of an environment can be illuminated with a flash of light and the reception signals reflected from any objects can be recorded with the receiving device.
  • Scanning lidar systems emit light beams that move in a scanning direction.
  • Point scanners illuminate areas of the surrounding area point by point.
  • Line scanners illuminate areas of the surrounding area line by line.
  • WO2021/249983A1 describes a lidar system in which a pulse sequence of laser light is emitted to scan the same area of an environment multiple times. The distances between the pulses of the pulse sequence differ from one another.
  • the lidar system can be designed as a point scanner, line scanner or flash lidar.
  • a method for operating a lidar system comprises the following steps:
  • an object By evaluating the received signal using the transmitted signal, an object can be detected in a detection area of the lidar system where the reflection occurred. For example, the distance to the object can be determined using a TOF method. It is also possible to determine a relative speed between the lidar system and the object.
  • the light pulse can be, for example, a short laser pulse that has a shape that is similar to a Dirac pulse.
  • the duration of the emitted real light pulse can be, for example, in the ns range.
  • the sequence of light pulses has the light pulses at short time intervals. The intervals can be, for example, in the ns range.
  • the light pulses of the sequence of light pulses are emitted in the same direction. This enables them to be reflected on the same object, whereby further information about the object can then be determined by reconstructing the received signal, e.g. the size and/or surface texture of the object.
  • Threshold sampling is a sampling method for a signal, particularly an analog signal, in which the sampling times are recorded at which the signal assumes the value of a predeterminable threshold.
  • the result of threshold sampling is therefore the sampling times that indicate the times at which the sampled signal assumes the threshold value.
  • the number of sampling times in a given period of time therefore depends on the signal shape of the sampled signal.
  • the method offers the advantage that the received signal and in particular the shape of the received signal can be efficiently reconstructed in a receiving device of a lidar system. From the reconstruction of the received signal, further information about the object on which the reflection occurred can then be obtained.
  • the iterative method comprises determining a cost function relating to a deviation of a model of the received signal from the threshold of the threshold sampling to the sampling points, wherein the model is iteratively improved and the cost function represents a measure of the quality of the model.
  • the iterative method can comprise the modeling of a single model.
  • the single model models a received signal for a single light pulse.
  • the respective individual models of the pulses of the sequence are arranged on the time axis in relation to the respective times of the transmission of the light pulses. This means that the individual models are arranged on the time axis in the same order and with the same distances from one another as the respective associated light pulses of the transmission signal.
  • a Superposition function is determined from a superposition of the individual models.
  • the superposition function can be, for example, an addition of the individual models or a convolution of the individual models.
  • the cost function is then determined for the deviation of the superposition function from the threshold of the threshold sampling at the sampling times.
  • the individual model has a first number of discrete support points and the threshold sampling has a second number of sampling times.
  • the support points of the individual model are those points at which the individual model has a function value assigned to a discrete value on the time axis.
  • the temporal distance between the support points can be the same, i.e. the support points can be equidistant. However, other temporal arrangements of the support points are also conceivable.
  • the first number is less than or equal to the second number.
  • the sampling times of the threshold sampling are determined by the function values of the received signal, namely when the received signal intersects the threshold value of the threshold sampling.
  • the number of support points of the individual model should now be at most as large as the number of sampling times for receiving the received signal. Otherwise, an under-determined problem could arise, the solution of which could be difficult.
  • the distances between the light pulses in the sequence can differ from one another and can become larger, in particular, over the course of the sequence. It is possible, for example, for the distances between the light pulses to form a geometric sequence.
  • the number of light pulses in the sequence can also vary and can be, for example, between 2 and 10, in particular 4, light pulses.
  • the distance between the light pulses and/or the number of light pulses in a sequence can be specified and, in particular, adjusted over the course of operation of the lidar system. In embodiments, it is possible to iteratively improve the distance between the light pulses and/or the number of light pulses in the sequence.
  • the sequence of light pulses can be sent out repeatedly. This means that after the transmission signal with the sequence of light pulses, another transmission signal with another sequence of light pulses can be sent out.
  • the sequences can be sent out in the same direction and/or in different directions, for example to achieve a scanning effect. It is also possible to send several sequences in one direction and then several sequences in the other direction. to send out signals in a different direction and thus create a scanning effect.
  • the distances between the sequences can remain the same or change over time. In particular, the distance between the sequences can be adjusted during the operation of the lidar system. In embodiments, it is possible to improve the distance between the sequences iteratively.
  • the cost function can be improved using a Monte Carlo method, in particular a Metropolis Monte Carlo method.
  • the threshold value of the threshold sampling, the shape of the individual model, the number of light pulses in the sequence, the distance between the light pulses in the sequence and/or the distance to the next sequence can be specified. These parameters can be improved iteratively in particular.
  • a machine learning method can be applied to the iterative improvement of the model.
  • the machine learning method can in particular relate to the choice of the threshold value, the shape of the individual model, the number of light pulses in the sequence, the spacing of the light pulses in the sequence and/or the spacing to the further sequence.
  • a lidar system has an optical transmitting device which is set up to transmit a transmitted signal with a sequence of light pulses.
  • the lidar system also has an optical receiving device which is set up to receive a reflection of the transmitted signal as a received signal.
  • the lidar system also has a computing unit which is set up to sample the received signal by means of threshold sampling and to reconstruct the received signal using the sampling times by means of an iterative method.
  • the receiving device can have a receiving sensor for opto-electrical conversion of the received signal and the computing unit can be designed for threshold scanning of the electrical signal output by the receiving sensor.
  • the receiving sensor can have at least one photodiode, in particular at least one avalanche photodiode, wherein the electrical signal output by the receiving sensor depends on a light intensity of the received signal received by the at least one photodiode.
  • the threshold value of the threshold scanning can depend on the receiving sensor, in particular the photodiode, of the receiving device.
  • a vehicle can have the lidar system described above, whereby the lidar system can be used, for example, for object detection, distance measurement, speed measurement and/or other properties of the detected object. Other properties of the detected object can, for example, relate to the size of the object.
  • Fig. 1 is a flow chart of a method for operating a lidar system
  • Fig. 2 a representation of a vehicle with Lidar system
  • Fig. 3 shows an example of an embodiment of a received signal as an electrical signal at the output of a photodiode
  • Fig. 4 shows another exemplary embodiment of the received signal as an electrical signal at the output of the photodiode
  • Fig. 5 a transmission signal with a transmitted light pulse and a reception signal with the corresponding reflected pulse
  • Fig 6 a transmission signal with a transmitted light pulse, a reception signal with the corresponding reflected pulse,
  • Fig. 7 a transmission signal with a sequence of light pulses, an associated reception signal, a threshold sampling of the reception signal and the associated sampling times
  • Fig. 8 the transmission signal with the sequence of light pulses, the respective associated individual models and a superposition function of the individual models with deviations
  • Fig. 9 the transmitted signal with the sequence of light pulses, improved individual models and an improved model of the received function with deviations.
  • FIG. 1 shows a method for operating a lidar system 10 (Fig. 2) with the steps:
  • Sl Transmission of a transmission signal 22 by an optical transmission device 12 of the lidar system 10.
  • the transmission signal 22 has a sequence of light pulses EP.1, EP.2, EP.3, EP4.
  • S2 Receiving a reflection of the transmitted signal 22 as a received signal 30 by an optical receiving device 14 of the lidar system 10.
  • S3 Sampling of the received signal 30 by a computing unit 16 of the lidar system 10 by means of threshold sampling.
  • S4 Reconstructing the received signal 30 by the computing unit 16 using sampling times of the threshold sampling by means of an iterative method.
  • FIG. 2 shows a schematic representation of a vehicle 20, for example a passenger car.
  • the lidar system 10 is arranged in a front area of the vehicle 20.
  • the lidar system 10 has the optical transmitting device 12 and the optical receiving device 14.
  • the transmitted signal 22 and the received signal 30 can be evaluated for the detection, distance determination and/or speed determination of an object O located in the detection area 26.
  • the computing unit 16 can also monitor and control the transmission process in the transmitting device 12 and the reception process in the receiving device 14.
  • the detection area 26 is located in front of the front area of the vehicle 20. In the example shown, this allows an area in the direction of travel in front of the vehicle 20 to be monitored. It is also possible to arrange the lidar system 10 in other areas of the vehicle 20, for example in the rear area and/or in side areas. It is also possible to arrange several lidar systems 10 on the vehicle 20, in particular in corner areas of the vehicle 20.
  • the lidar system 10 can detect stationary or moving objects O, in particular vehicles, persons, animals, plants, obstacles, road irregularities, in particular potholes or stones, road markings, traffic signs, spaces, in particular parking spaces, precipitation or the like, are detected in the detection area 26.
  • the method for operating the lidar system 10 enables the determination of further properties of the object by reconstructing the received signal 30, which go beyond detection, distance determination or speed determination.
  • the lidar system 10 can determine the distance to a reflective, remitting object O using the ToF principle.
  • the lidar system 10 can emit short laser pulses EP.1, EP.2, EP.3, EP4, e.g. of nanosecond duration, into its surroundings, in particular its detection area 26, via the transmitting device 12.
  • the lidar system 10 can receive the returning light reflected from the object O via its receiving device 14. The time between emission and reception of the light is measured with high accuracy (gigahertz), and the light travel time from the lidar system 10 to the object O and back to the lidar system 10 is multiplied by half the speed of light to calculate the distance between the lidar system 10 and the detected object O.
  • Figure 3 shows a representation of the received signal 30 in the form of the electrical output signal of an avalanche photodiode and its threshold value TDC for time-to-digital conversion.
  • the threshold value TDC defines the width EPW of the peak of the received signal 30 via a threshold sampling.
  • the echo pulse width in meters resulting from the width EPW of the peak is the product of the speed of light c and the temporal width EPW of the peak of the received signal 30 sampled with the threshold value TDC.
  • the temporal width EPW is the result of the distance between the two sampling times at which the received signal 30 intersects the threshold TDC.
  • the reflected light is received as a reception signal 30 by one or more reception sensors, such as avalanche photodiodes.
  • the at least one reception sensor converts a momentary light intensity of the reception signal 30 into an electrical current and outputs an electrical signal corresponding to the reception signal 30.
  • the electrical signal output by the reception sensor corresponds to a current that varies over time. It is shown in Figure 3 as a reception signal 30 with a pulse height PH.
  • the light pulse received by the receiving sensor has a duration of a few, approximately 10 nanoseconds (ns) and is converted into a received signal 30 is translated into the form of an electrical current signal which rises above a noise floor NF at a time of approximately 300 ns before falling back towards the noise floor NF at approximately 310 ns.
  • the final divisor 2 takes into account the light path from the lidar system 10 to the object 0 and back.
  • the peak shape of the received signal 30 in the form of the electrical signal over time in Fig. 3 follows the temporal distribution of the light intensity of the reflected transmitted signal 22 if effects of the photodiode such as saturation and non-linearity can be neglected.
  • the temporal distribution of the received signal 30 depends on various parameters. These parameters include, for example, the shape and duration of the emitted laser pulse EP.l, EP.2, EP.3, EP4, the degree of atmospheric scattering and the geometric and material properties of the object O from which the light pulse EP.l, EP.2, EP.3, EP4 was reflected to the lidar system. Blunt objects 0 with irregular surfaces typically "wash out” and increase the temporal duration of the received signal 30, while highly reflective objects such as retroreflectors tend to produce a clearly defined return peak of short duration and pronounced height PH.
  • the shape of the received signal 30 includes information about the properties of the detected object O, it may be desirable to know the entire shape of the received signal 30 and to transmit information about the signal shape to electronic signal processing methods and object recognition methods and/or classification methods.
  • sampling the entire received signal 30 may be costly in terms of hardware, e.g. in application-specific integrated circuits (ASICs), and in terms of the amount of data generated that then has to be further processed.
  • ASICs application-specific integrated circuits
  • the maximum pulse height PH which correlates with the peak intensity of the received signal 30, can be determined.
  • This peak intensity in turn correlates with the optical reflectivity and other properties of the reflective object. object 0.
  • the peak height of the received signal 30 therefore correlates with the reflectivity of the object 0 and can additionally be used for object classification, for example.
  • the width of the received signal 30 can be sampled.
  • a threshold value TDC for the time-to-digital conversion is applied to the received signal 30.
  • the time tl at which the received signal 30 first rises above the threshold value TDC (approx. 301 ns in Fig. 3) and the later time t2 at which the received signal 30 falls below the threshold value TDC (approx. 308 ns in Fig. 3) are both determined in the computing unit 16, e.g. in an ASIC of the computing unit 16.
  • tl is a first approximation for the round trip time of light in calculating the distance of the object O.
  • the difference or "temporal width", t2 - tl, of the peak can be multiplied by the speed of light to obtain the so-called Echo Pulse Width EPW, see Fig. 3.
  • the width EPW for the received signal 30 is contained in a point cloud output of the lidar system 10.
  • Highly reflective objects O such as the car's license plate and a traffic sign, typically result in "wide" received signals 30 with large width EPW.
  • the width EPW can be used to classify the reflectance of an object, alternatively or in addition to the pulse height PH.
  • a technical and economic advantage of sampling the width EPW instead of the height PH is a simpler computing unit 16, e.g. the ASIC logic and/or the ASIC design. Sampling the width requires a direct sampling, i.e. comparing the instantaneous signal value of the received signal 30 with the threshold value TDC. The resulting logical 0/1 output corresponds to a signal below or above the threshold value TDC.
  • Sampling the peak height PH requires a somewhat more complicated (numerical) differentiation of the received signal 30 in order to determine the point in time at which the derivative changes sign and the peak PH is reached.
  • width EPW has the advantage of being simple and economical to carry out, it can lead to problems in the comparability between point cloud data from width EPW sampling and height PH sampling.
  • Certain object detection and classification methods, such as deep neural networks, that work with the point cloud as input might be designed to use width EPW sampling or height PH sampling. Choosing the other sampling may lead to undesirable effects.
  • Figure 4 shows another example of a special effect of scanning the width EPW.
  • the spread geometric layer projection corresponds to an extended range of object distances from which the light returns to the lidar system 10 for a single emitted light pulse 28.
  • This extended range of illuminated and reflected object distances leads to a temporally scattered received light intensity and a correspondingly scattered received signal 30 in Fig. 4.
  • the shape of the received signal 30 then differs significantly from the typical shape for an object 10 that is illuminated at an obtuse angle. With a grazing incidence of the transmitted signal, the correlation between intensity, i.e. height PH, and width EPW can therefore be lost.
  • the method described below for operating the lidar system 10 can be used.
  • the method of operating the lidar system 10 described below enables a good reconstruction or approximation of the shape of the reflected returning pulse RP to be found on a manageable finite set of sample points 32.
  • the number of sample points 32 is greater than 1, as in the case of sampling the height PH, and also greater than 2, as in the case of sampling the width EPW.
  • Figure 5 shows a schematic, single short (nanosecond or shorter) emitted light pulse EP in the upper part.
  • Figure 5 shows the intensity of the returning, reflected pulse RP as a function of time in the lower part.
  • the upper part of Figure 6 shows a schematic representation of a very short (nanosecond or shorter) emitted laser pulse and the resulting reflected pulse RP.
  • the representation in the lower part of Figure 6 corresponds to the electrical signal at the output of the photodiode of the receiving unit 14.
  • an approximate value in the form of an individual model 28 is used for the reflected pulse R.P.
  • a reasonable candidate for such a first approximation in the form of the individual model 28 is, for example, a Gaussian bell curve with estimated height and width.
  • the number of support points for the approximation is 6 for the embodiment shown in Figure 6.
  • the number of support points is preferably chosen to be equal to or smaller than the number of sampling points of the threshold sampling of Figure 7. If the number of support points for the discrete individual model 28 is chosen to be larger than the number of sampling times determined by the threshold sampling, problems can arise. An underdetermined optimization problem could arise for which a solution would be difficult.
  • the support points for the individual model 28 can be arranged equidistantly in time in a time grid ZR.
  • an equidistant temporal spacing of the support points is optional.
  • Other temporal arrangements of the support points of the individual model 28 are also conceivable.
  • the lower part shows the (unknown) shape of the curve of the reflected pulse RP, which corresponds to the emitted laser pulse EP shown at the top of Figure 6.
  • a time grid ZR which here consists of 6 times.
  • the first estimate (1st approximation) for the individual model 28 includes 6 support points at the times of the time grid ZR with values that can correspond to a reasonable approximation of a reflected light pulse RP, e.g. a Gaussian bell function.
  • This first approximation value of the individual model 28 represents the initial state of an iterative process. Using the iterative process, e.g. a Monte Carlo sampling process, the true shape of the reflected pulse RP can be approximated.
  • Figure 7a schematically shows a sequence of four very short, emitted laser pulses EP.1, EP.2, EP.3, EP.4 with time intervals in a geometric sequence of 1 ns, 2 ns, 4 ns.
  • the optical transmission device 12 emits, e.g. via a laser diode, a sequence, also called cadence, of light pulses EP.1, EP.2, EP.3, EP.4.
  • the light pulses EPI, EP.2, EP.3, EP.4 have a geometric time profile.
  • Geometric time course also called geometric progression, is understood here to mean an increasing time interval between the emitted light pulses EP1, EP2, EP3, EP4, whereby the time interval between pulse number n and pulse number n + 1 exceeds the interval between n - 1 and n by a constant factor.
  • the geometric progression factor does not have to be 2, but can take on any numerical value greater than 1. The exact value of the progression factor can, for example, be adjusted during a system test of the lidar system and/or improved as part of the iterative process for operating the lidar system 10.
  • Figure 7b shows the intensity of the light of the received signal 30 coming back from an object O illuminated by the four pulses of Fig. 7a. It is a superposition, e.g. the sum, over the four individual reflected pulses RP. A threshold value TDC is applied to the total received intensity of the received signal 30.
  • Figure 7b thus shows the received signal 30 of the reflected light, i.e. the reflected sequence of light pulses returning from an illuminated object O and arriving at the receiving unit 14.
  • the received signal 30 is a superposition, in the example shown a sum, of the individual reflected pulses RP for each emitted light pulse EP.1, EP.2, EP.3, EP.4 in the sequence. Examples of such individual reflected pulses RP are shown, for example, in the lower part of Figure 5, in the lower part of Figure 6 and in Figure 7b.
  • the received signal 30 thus corresponds to the superposition of the reflected pulses RP. In the embodiment shown in Figure 7b, the superposition and thus the received signal 30 has several local minima and maxima.
  • the pattern of the minima and maxima in the received signal 30 would repeat itself almost perfectly. In such a case, the shape of the received signal 30 would allow little or no additional information to be determined that is not already contained in the minima and maxima generated by two transmitted pulses EP.1, EP.2, EP.3, EP.4.
  • each pulse reflection i.e. each reflected pulse R.P.
  • Figure 7c shows six sampling points. They mark the sampling times at which the intensity signal of the received signal 30 has intersected the threshold value TDC.
  • the threshold value TDC is compared with the received signal 30.
  • the comparison is carried out using a robust and cost-effective sampling logic, as can also be used for methods described with reference to Figure 2 and/or Figure 3.
  • the sampling logic can be implemented using a level comparison logic, which can be easily implemented, for example, in an ASIC or another embedded integrated circuit.
  • the six resulting sampling points 32 are shown in Figures 7c and 7d.
  • the sampling times are at the times at which the received signal 30 exceeds or falls below the threshold value TDC threshold. Note that the first local minimum of the received signal 30 does not fall below the threshold value TDC, while the second local minimum only just falls below the threshold value TDC. If these minima were slightly more or less pronounced, a different number of sampling points would have resulted, e.g. 8 or 4, instead of 6.
  • the set of digitized sampling times shown in Figure 7d contains the information that can be used in further process steps, e.g. by a signal processing processor, to reconstruct the shape of the received signal.
  • the number of sampling times can vary: If the first local minimum of the intensity curve of the received signal in 7b and 7c had been somewhat more pronounced, then there would have been 8 sampling times instead of 6. If the second local minimum in 7b and 7c had been somewhat less pronounced, then there would have been 4 sampling times instead of 6.
  • Figures 7a and 7b illustrate further information which is passed on to the signal processing processor on which the method can be carried out and is thus available for the iterative method.
  • This information includes the sequence of light pulses EP.1, EP.2, EP.3, EP.4, as shown in Figure 7a, for example, and the number of sampling points, which is 6 in the case shown in Figure 7, and the threshold value TDC, which corresponds to the height of the sampling points 32 along the vertical axis.
  • the height of the sampling points 32 is the same for each of the sampling points 32.
  • the information which is passed on to the processor can also include the time for each of the sampling points 32, i.e. the sampling times.
  • the sampling times correspond to the value along the horizontal axis in Figure 7d.
  • the proposed sampling is reduced to the sampling of the width EPW, which was described in connection with Figure 3 and Figure 4.
  • a method for reconstructing the received signal after threshold sampling now has the following:
  • a single model 28 ( Figure 6) of a reflected pulse RP is calculated for the reflected pulse to be reconstructed.
  • a reasonable candidate for this first approximation is, for example, a Gaussian bell curve with an estimated height PH and width EPW.
  • the number of support points for the approximation in the single model 28 is 6 in the case of Figure 6.
  • the number of support points should be chosen to be equal to or smaller than the number of sampling times of the threshold sampling.
  • the number of sampling times corresponds to the number of sampling points 32 of the threshold sampling and is also 6 in the example shown in Figure 7. If the number of support points for the discrete single model 28 is chosen to be greater than the number of sampling points 32 , then an underdetermined optimization problem could arise that might be unsolvable.
  • the support points for the individual model 28 can be arranged equidistantly in time, as suggested in Figure 6.
  • an equidistant time interval is optional and can be replaced by another, possibly more suitable, time grid ZR.
  • Figure 8 shows the construction of a cost function with a non-negative, scalar value.
  • an individual model 28.1, 28.2, 28.3, 28.4 is created to approximate the reflected pulse R.P and shifted in time so that it is located around the corresponding transmitted light pulse EP.1, EP.2, EP.3, EP.4.
  • the superposition e.g. sum of all of these shifted individual models 28.1, 28.2, 28.3, 28.4, is determined.
  • the resulting sum curve 34 is the first approximation for the reconstruction of the received signal 30.
  • the shape of the individual model 28 is copied and shifted in time to each point in time at which a light pulse EP.1, EP.2, EP.3, EP.4 is emitted.
  • the sum of the shifted individual models results in the first approximation 34 for the received signal 30.
  • four individual models 28.1, 28.2, 28.3, 28.4 are added and linear interpolation lines are drawn between the support points of the individual models 28.1, 28.2, 28.3, 28.4 and the sampling points 32.
  • the approximation 34 deviates from the threshold value TDC by Dl, D2 ... D6. In the case shown in Figure 8, both D5 and D6 are almost zero.
  • the cost function which represents a quality for the iterative process and which can be minimized in particular for optimization, can be the sum of all squared deviations, ie Dl 2 + D2 2 + D3 2 + D4 2 + D5 2 + D6 2 .
  • the individual model 28, 28.1, 28.2, 28.3, 28.4 can be iterated in an iterative procedure, e.g. a Metropolis Monte Carlo algorithm to convergence.
  • the cost function can be minimized in the following way.
  • step j one of the support points of the single model 28 is randomly selected and the intensity value (along the vertical axis in Figure 7) of the support point is randomly changed by adding or subtracting a random number.
  • the cost function costj corresponding to the newly modified single model is calculated, and the Boltzmann weight Bj is calculated as follows
  • T a dimensionless, non-negative "temperature” that can be chosen to be fixed or optionally varied as a function of the iteration index j in an optimization procedure known as "simulated annealing". If T is chosen to be large for small values of j and then gradually reduced to a value of (nearly) zero as j increases, then this "annealing" usually helps to avoid the optimizer getting stuck in local minima of the cost function.
  • the random shift of a sample point in step j is calculated with a probability
  • a random number q is drawn from a uniform distribution on the interval [0, 1), and the move is accepted if q ⁇ pj. Otherwise, the Monte Carlo move is discarded and the sample point is placed back at the position where it was in iteration step j-1.
  • Figure 9 shows the reconstruction 34.
  • M of the received signal 30 after the iterative method has converged and found an approximate value 28.1.
  • the converged single model 28.1.M, 28.2.M, 28.3.M, 28.4.M is close to the (unknown) true shape of the reflected pulse RP.
  • the sum 34.M over the converged shifted single models 28.1.M, 28.2.M, 28.3.M, 28.4.M is close to the (unknown) true shape of the received signal 30.
  • the value of the converged cost function will be low, although not zero, due to the imperfections e.g. due to the sampling, discretization and/or linear interpolation.
  • the method can therefore be used to determine discrete values of the converged approximation function 34.M of the received signal 30.
  • the form of this approximation function 34.M can include information that can go beyond the information of a simple received pulse, such as that described in connection with Figure 3 or Figure 4.
  • the number of sampling points 32 depends on the number of minima and maxima in the received signal 30, which in turn depends on the number of pulses EP emitted in the cadence, i.e. the sequence of light pulses in the transmitted signal 22. As the number of light pulses EP emitted increases, the number of sampling points 32 also generally increases.
  • a well-resolved approximation 34. M with many sample points facilitates object detection and classification by downstream methods.
  • the iterative method described above e.g., the Metropolis Monte Carlo method, can be computationally expensive, potentially causing problems for use on embedded hardware, such as an electronic control unit on board a vehicle, inside or outside the lidar system 10.
  • the method processes, e.g. concerning the sequence of light pulses, concerning the threshold value TDC, concerning the repetition of the sequences and/or concerning the sampling.
  • the converged results of the method in particular the converged individual model 28, can then be used for training an artificial neural network.
  • the input layers of this network receive the input data, and the output layer can represent a channel for each sampling point 32 of the individual model 28.
  • the neural network can be trained with algorithms such as backpropagation if a sufficient amount of training data, i.e. input-output pairs, is available.
  • the training data can be obtained, for example, during test drives in which the lidar system 10 is confronted with different objects 0 and different environmental conditions.
  • test drives can be carried out under different climatic conditions and/or with different vehicles 20. In this way, the variations in the training data can be increased and the performance of the neural network can be improved under conditions outside the training set.
  • iterative procedures can be applied to some of the input data that a lidar system 10 receives during operation.
  • the resulting new result of the iterative procedure e.g. the resulting Metropolis Monte Carlo output
  • Such unsupervised learning of the network can improve and stabilize the performance of the network under slowly varying (drifting) environmental conditions.
  • Feedforward through a trained neural network of the expected dimension with e.g. 10 or fewer input channels, 10 or fewer output channels and few, e.g. 2-5, layers is computationally cheap and is therefore suitable for integration into embedded hardware, such as in the vehicle.
  • the approximation of arbitrarily short, Dirac-delta-like, transmitted light pulses EP was used.
  • the transmitted light pulses EP are not much shorter than the returning intensity signal, the previously described summations for forming the superposition function can be replaced by a convolution.
  • the principle of the method and the functioning of the proposed lidar system 10 remains unaffected by using the convolution instead of the sum and offers results in an equally advantageous manner.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Electromagnetism (AREA)
  • Optical Radar Systems And Details Thereof (AREA)

Abstract

L'invention concerne un procédé de fonctionnement d'un système lidar, comprenant les étapes consistant à : - émettre un signal de transmission (22) avec une séquence d'impulsions lumineuses (EP.1, EP.2, EP.3, EP4), - recevoir une réflexion du signal de transmission (22) en tant que signal reçu, balayer le signal reçu à l'aide d'un processus de balayage de seuil, et - reconstruire le signal reçu à l'aide des temps de balayage du processus de balayage de seuil au moyen d'un procédé itératif. Pour chaque impulsion lumineuse (EP.1, EP.2, EP.3, EP.4) dans la séquence, un modèle individuel (28.1, 28,2, 28,3, 28,4) est de préférence généré pour approximer l'impulsion réfléchie et est décalé dans le temps de sorte que le modèle est situé autour de l'impulsion lumineuse de transmission correspondante. La superposition, par exemple la somme, de tous les modèles individuels décalés dans le temps est ensuite déterminée. La courbe de somme ainsi obtenue (34) est la première approximation pour reconstruire le signal reçu. Afin de calculer la courbe de somme (34), l'invention propose une interpolation linéaire des modèles individuels décalés aux instants coïncidant avec les temps de balayage (32). Une fonction de coût peut être définie comme la somme des différentiels au carré (D1, D2, D3, D4, D5, D6) entre la valeur de seuil de TDC et l'approximation interpolée linéairement (34) du modèle du signal reçu. La forme du modèle individuel (28) est ainsi copiée et décalée dans le temps à chaque instant d'une émission d'impulsions lumineuses. Les modèles individuels peuvent être itérés dans un modèle itératif, par exemple un algorithme Metropolis Monte Carlo, pour un processus de convergence. L'invention concerne en outre un système lidar, un véhicule comprenant un tel système lidar, et l'utilisation du système lidar dans un véhicule.
PCT/EP2023/083270 2022-12-01 2023-11-28 Procédé de fonctionnement d'un système lidar, et système lidar WO2024115439A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
DE102022131848.6A DE102022131848A1 (de) 2022-12-01 2022-12-01 Verfahren zum betrieb eines lidar-systems und lidar-system
DE102022131848.6 2022-12-01

Publications (1)

Publication Number Publication Date
WO2024115439A1 true WO2024115439A1 (fr) 2024-06-06

Family

ID=89029871

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2023/083270 WO2024115439A1 (fr) 2022-12-01 2023-11-28 Procédé de fonctionnement d'un système lidar, et système lidar

Country Status (2)

Country Link
DE (1) DE102022131848A1 (fr)
WO (1) WO2024115439A1 (fr)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170329010A1 (en) * 2016-05-10 2017-11-16 Texas Instruments Incorporated Methods and apparatus for lidar operation with pulse position modulation
US20180284229A1 (en) * 2017-03-29 2018-10-04 SZ DJI Technology Co., Ltd. Light detecting and ranging (lidar) signal processing circuitry
US20200284907A1 (en) * 2019-03-08 2020-09-10 Wisconsin Alumni Research Foundation Systems, methods, and media for single photon depth imaging with improved precision in ambient light
WO2021249983A1 (fr) 2020-06-10 2021-12-16 Robert Bosch Gmbh Procédé et dispositif pour garantir un domaine de non-ambiguité d'un capteur lidar et capteur lidar de ce type
US20220317267A1 (en) * 2021-04-02 2022-10-06 Luminar, Llc Reconstruction of pulsed signals

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102011056963C5 (de) 2011-12-23 2018-03-01 Sick Ag Messung von Entfernungen nach dem Signallaufzeitprinzip
US20230057064A1 (en) 2020-01-24 2023-02-23 Outsight A laser detection and ranging (lidar) device
DE102021101790A1 (de) 2021-01-27 2022-07-28 Osram Gmbh Detektionssystem und verfahren dafür

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170329010A1 (en) * 2016-05-10 2017-11-16 Texas Instruments Incorporated Methods and apparatus for lidar operation with pulse position modulation
US20180284229A1 (en) * 2017-03-29 2018-10-04 SZ DJI Technology Co., Ltd. Light detecting and ranging (lidar) signal processing circuitry
US20200284907A1 (en) * 2019-03-08 2020-09-10 Wisconsin Alumni Research Foundation Systems, methods, and media for single photon depth imaging with improved precision in ambient light
WO2021249983A1 (fr) 2020-06-10 2021-12-16 Robert Bosch Gmbh Procédé et dispositif pour garantir un domaine de non-ambiguité d'un capteur lidar et capteur lidar de ce type
US20220317267A1 (en) * 2021-04-02 2022-10-06 Luminar, Llc Reconstruction of pulsed signals

Also Published As

Publication number Publication date
DE102022131848A1 (de) 2024-06-06

Similar Documents

Publication Publication Date Title
EP2541273B1 (fr) Détection et détermination de distance d'objets
EP2823331B1 (fr) Procédé et dispositif pour appréhender des objets dans l'environnement d'un véhicule
EP2917756B1 (fr) Dispositif de détection optoélectronique à balayage à seuil de détection, véhicule automobile et procédé afférent
DE102018201220A1 (de) Abstandsdetektionssystem, Verfahren für ein Abstandsdetektionssystem und Fahrzeug
DE19717399A1 (de) Einrichtung zur Bestimmung von Abstand und Art von Objekten sowie der Sichtweite
DE102005013555A1 (de) Radarvorrichtung
DE19708014A1 (de) Vorrichtung und Verfahren zum Erfassen eines Objekts in einem vorgegebenen Raumbereich, insbesondere von Fahrzeugen für die Verkehrsüberwachung
DE102017119042A1 (de) Vermeidung von Totwinkelwarnungen durch Gischt
DE102019214612A1 (de) Verfahren und Steuergerät zum Erkennen eines Objekts in einer Umgebung eines Fahrzeugs
DE102016225494A1 (de) Verfahren und vorrichtung zum erfassen eines zielobjekts
WO2003036320A2 (fr) Procede et dispositif de mesure des vitesses d'objets en mouvement
DE102020115997A1 (de) Fahrzeugsensorverbesserungen
DE102019205565A1 (de) Verfahren und Vorrichtung zum Bewerten einer Objekthöhe mittels von einem an einem Fahrzeug angebrachten Ultraschallsensor empfangenen Ultraschallsignalen
EP2159603A1 (fr) Procédé de détection d'objets et capteur de détection d'objets
EP3663881B1 (fr) Procédé de commande d'un véhicule autonome en fonction des vecteurs de mouvement estimés
EP2766742B1 (fr) Procédé et dispositif de mesure optique de distances sur de grandes étendues
EP1085346B1 (fr) Procédé pour mesurer la distance d'un objet, de particules atmospheriques et similaires utilisant des signaux de Lidar ou Radar à laser
EP3867666B1 (fr) Procédé de détection d'au moins des compositions de particules dans une zone de surveillance comprenant un dispositif de détection optique et dispositif de détection
DE102020110809B3 (de) Verfahren und Vorrichtung zum Erkennen von Blooming in einer Lidarmessung
WO2024115439A1 (fr) Procédé de fonctionnement d'un système lidar, et système lidar
DE10213901A1 (de) Verfahren zur Messung der Relativgeschwindigkeit eines Objekts
WO2019101506A1 (fr) Procédé de fonctionnement d'un capteur lidar et capteur lidar
DE19704220A1 (de) Verfahren und Vorrichtung zum Bestimmen eines Abstandes zwischen Fahrzeug und Hindernis
DE102021117361A1 (de) Verfahren zum Betreiben einer optischen Detektionsvorrichtung, Detektionsvorrichtung und Fahrzeug mit wenigstens einer Detektionsvorrichtung
DE10149423B4 (de) Verfahren und Vorrichtung zur Messung von Entfernungen in optisch trüben Medien