WO1995007473A1

WO1995007473A1 - Method and device for detecting and locating obstacles in a vehicle environment

Info

Publication number: WO1995007473A1
Application number: PCT/FR1994/001043
Authority: WO
Inventors: Omer Mercier; Michel Trouble
Original assignee: Framatome
Priority date: 1993-09-10
Filing date: 1994-09-06
Publication date: 1995-03-16
Also published as: FR2709834B1; EP0717854A1; FR2709834A1

Abstract

The invention relates to a method for detecting and locating obstacles in a vehicle environment. To this effect, it comprises the steps of emitting signals in the direction of said environment from a plurality of radars (2) on the vehicle, deducing from the echo signals received the probability of occupancy, by an obstacle, of a set of elementary cells (9) of said environment, and combining the probabilities relative to each of said cells.

Description

Method and device for detecting and locating obstacles in the environment of a vehicle

The present invention relates to a method and a device for detecting and locating obstacles located in the environment of a stopped or moving vehicle.

An important function of mobile robotics concerns the detection of objects or obstacles located in front of or on the side of the carrier vehicle, insofar as these can impede its movement and must be bypassed, or require its complete stop for the sake of safety. or, on the contrary, are sufficiently distant or in a direction such that the vehicle can continue its way without deviation.

A secondary aspect of this detection consists in using a priori knowledge of the (absolute) positioning of the detected object with respect to a map or a larger site, so that by detecting this object and by measuring its coordinates (site and distance ) information about the absolute position of the vehicle itself can be obtained.

A convenient means of detecting objects located in the periphery of a vehicle and making it possible to measure the coordinates thereof is by using ultrasonic sensors.

According to this well-proven technology, an acoustic wave is emitted periodically and, being reflected by the objects constituting obstacles to its propagation, returns to the sensor after a delay time, which depends directly on the distance of the object thus detected.

To ensure peripheral detection of a vehicle, it is necessary to have a sufficient number of such sensors around its periphery as a function of the opening angle of the elementary sensor. This number is even higher than the sensors will have a narrow detection lobe.

The use of these sensors provides distance information relative to an obstacle, but the angular orientation information is all the less good. that the sensors are wide lobe. Indeed, insofar as a sensor sees a given obstacle not at a point in space but in an ambiguous way at all points located at equal distance, if we combine several sensors to ensure coverage of the perimeter of the vehicle, the indeterminacy of the angular position of the detected obstacles can be excessively large.

Conversely, finer detection requires a greater number of elementary sensors with narrower lobes, which increases the overall cost of detection.

Another drawback of the use of ultrasonic sensors is their very high sensitivity to external acoustic disturbances, to anomalies in reflection on certain obstacles, as well as the sensitivity to wind of sound waves, the echo can even be blown and not never go back to its source.

The present invention aims to overcome these drawbacks.

To this end, the invention firstly relates to a method for detecting and locating obstacles located in the environment of a vehicle, characterized in that it comprises the steps consisting in transmitting signals towards of said environment from a plurality of radars arranged on the vehicle, to deduce from the signals received in echo the probabilities of occupation by an obstacle of a set of elementary cells of said environment, and to merge the probabilities relating to each of said cells .

Information fusion algorithms are already known in the art. However, the fact of merging the probabilities of occupation of elementary cells, these probabilities being themselves obtained from radar signals, can give, with the technique according to the invention, results quite surprising in their precision, even allowing the mapping of obstacles detected and therefore making their avoidance possible as well as the location of the vehicle itself, due to the high quality of the environment modeling thus carried out.

In addition, unlike ultrasonic sensors, radars can be used outdoors and in any weather. Finally, it has been observed that the invention even makes it possible to use radars of relatively poor individual performance without altering the results of the global modeling.

In one embodiment of the invention, the fusion is carried out on the values of the probabilities provided by a plurality of radars, and in particular two radars.

In this case, the fusion is carried out spatially so as to associate data returned simultaneously by a subset of radars having detected the same obstacle from different angles.

In another embodiment, the merge is performed on the value of the probability supplied by a radar and the value of the probability previously known for the same cell.

In this case, the fusion is carried out temporally, the same obstacle being seen successively by the same sensor which will have advanced at the same time as the vehicle. In this case, the merge operation is applied to the value already assigned to the cell.

It is possible to deduce from the signals received in echo distance information and angular information on the obstacles and to determine said probabilities of occupation from this information.

Preferably, a step of thresholding the results of the fusion is provided so as to obtain a stripped map of the various artefacts which may come from specular or multiple reflection, from the emitted radar waves.

According to a particular embodiment of the invention, the signal emitted by the radars is frequency modulated, the signal received in echo is mixed with a fraction of the signal emitted so, after elimination of the carrier frequency, to form a beat signal, and the beat pulses are counted during a predetermined interval to provide the distance from an obstacle. Alternatively, this distance can be provided by Fourrier transformation.

Advantageously, the emission of each radar lasts a limited time and the repetition logic can even be coded for the purpose of identifying the source.

The invention also relates to a device for modeling the environment of a vehicle, characterized in that it comprises a plurality of radars mounted on the vehicle and a processing unit arranged to merge the data provided by the radars.

Thanks to the fusion operation, it is possible without disadvantage to choose radars with a large lobe, therefore of average performance and of low cost.

It will be observed that radars of this type, in particular the current Doppler radars, are generally used in speed measurement whereas here, it is the distance which is measured principally, the speed measurement being possibly used only on an ancillary basis. In addition, while the radars used for distance measurements are generally radars operating in pulses, radars with continuous emission and operating in frequency modulation are preferably used here.

A particular embodiment of the invention will now be described by way of nonlimiting example, with reference to the appended schematic drawings in which:

FIG. 1 is a schematic top view of a vehicle fitted with the means of the invention,

FIG. 2 is a functional diagram of these means,

FIG. 3 illustrates the signals emitted by the radars,

FIG. 4 represents the modeling of the environment of the vehicle, FIG. 5 illustrates the function of distribution of the probabilities of occupation of a cell by an obstacle as a function of the distance from the focus of an antenna, and

- Figure 6 illustrates the function of distribution of the probabilities of occupation of this cell according to the angular difference between its direction and the axis of the antenna.

FIG. 1 represents an automatic vehicle provided with a platform 1 on which are mounted a plurality of radars 2, for example twenty. The radars 2 are here wide-lobe radars, that is to say with an opening angle 2Θ typically between 15 ° and 120 °. Remember that a large angle 2Θ is sought to obtain good peripheral coverage around the vehicle with as few sensors as possible.

The radars 2 are here FMCW radars with frequency modulation and continuous emission or by time window.

The radars 2 are placed at the periphery of the vehicle so as to obtain a vision as complete as desired of its environment. The smoothness of detection coverage and modeling in general, directly depends on the number of radars used, the individual lobes of which can be joined arbitrarily, as shown forwards (arrow 3) and the side of the vehicle.

We see in Figure 2 three of the radar antennas 2a, 2b and 2c. Each of these antennas is connected via an interface circuit 4a, 4b, 4c respectively, to a unit 5 capable of producing the trigger signals Sa, Sb and Se respectively of the radars. Alternatively, the trigger signals may be common to all radars.

The signals Sa, Sb, Se are frequency modulated as shown in FIG. 3. A carrier frequency Fp is modulated by a sawtooth wave of frequency Δf. This frequency generation can be ensured at unit 5 by a GUNN oscillating diode. Optionally, for each signal Sa, Sb, Se, the modulated transmission lasts a limited time D. This thus facilitates the differentiation of the signals received as well as the extraction of the information sought. Furthermore, by coding the repetition logic of the transmissions, that is to say the duration T in FIG. 3, for example according to a pseudo-random model, it is possible to personalize each of the transmissions. It is thus in particular possible to operate several vehicles in the same environment.

The signal received in return by each antenna 2a, 2b, 2c is processed in an analog processing unit 6a, 6b, 6c respectively.

The received signal is out of phase with the transmitted signal. By mixing in the corresponding unit 6 with a fraction of the transmitted signal, after elimination of the carrier frequency, a beat signal at a typical frequency of the order of 1 kHz is obtained.

The signal is first filtered to eliminate the high frequencies, corresponding to the most distant obstacles. We thus retain only the signal backscattered by the first obstacle encountered, which is then processed by Fourrier transform.

Furthermore, the comparison of the frequency of the backscattered carrier wave and that of the carrier wave of the transmitted signal gives the Doppler frequency F ^ which is proportional to the speed of approaching or moving away from the vehicle of the detected obstacle. , depending on the relationship:

v = 1/2 λ F _d

where λ = c / F (propagation speed / emission frequency)

The speed information is not necessary for the implementation of the invention. Its availability is nevertheless an advantage in applications.

The digitized distance and speed information is supplied by the analog processing units 6a, 6b, 6c to the digital processor 7. The processor 7 essentially consists of a calculation unit capable of accessing a memory 8.

Each element of the memory 8 is associated with an elementary cell 9 of the environment of the vehicle. The content of each of its memory elements represents the probability that a particular cell, identified by its site φ and its distance d from the vehicle, contains an obstacle.

FIG. 4 represents 2 radars 11 and 12 each having detected an obstacle, the first at the distance di and the second at the distance d2- By convention, it is decided that the probability of occupation of a cell by an obstacle is between -1 (the cell is certainly not occupied) and +1 (the cell is certainly occupied), the probability p = 0 corresponding to the uncertainty.

Knowing the position of the focal point of each antenna as well as its line of sight, a probability of occupation is assigned to each cell as a function of the distance at which an obstacle has been detected, of the distance from this cell to the focal point of the antenna and the angular difference between the line of sight of the antenna and the direction of the cell. By way of example, FIG. 5 represents the function of distribution of the probabilities of occupation of a cell as a function of its distance from the focus of the antenna of the radar 11, and FIG. 6 the function of distribution of the probabilities as a function the angular difference α between the axis of this antenna and the direction considered. The maximum probability is of course encountered for the distance d-j and in the direction of sight of the antenna.

The different probability zones are shown in FIGS. 4 to 6. Zone B of positive probability in FIG. 5 corresponds to two annular bands on either side of the circle with radius dj. It is also verified in FIG. 5 that only the first obstacle encountered is taken into account since beyond a certain distance (zone D) function of d "| no information is provided by the measurement. Similarly in FIG. 6, no information is provided by the measurement outside the emission lobe (zone A). The possible levels of quantification between the values -1 and +1 depend on the performance of the computing equipment used, the overall performance of the representation of the environment being all the better as these levels are more numerous.

When a cell has been assigned in this way to a given cell, by the measurement carried out using the radar 11, a probability p1 of occupation of this cell by an obstacle and, by the measurement carried out using the radar 12, a probability p2, the following fusion algorithm is implemented:

- if p1> 0 and p2> 0.

F (p1, p2) = p1 + p2 - p1.p2

- if p1 <0 and p2> 0

F (p1, p2) = p1 + p2 + plp2

if p1 <0 and p2 <0

p1 + p2

F (p1, p2) =

1 - Min (p1, p2)

This algorithm can be implemented in the manner described above, that is to say spatially, by merging the probability information coming from two radars, or even in a temporal manner, by using the probability information coming from 'a single speed camera and merging it with the probability already present for the cell considered in memory 8.

The probabilities of occupying the space are continuously expressed between -1 (cell certainly empty) and +1 (cell certainly occupied). A thresholding of the results of the raw fusion is then carried out so as to obtain a map of the environment of the vehicle on which the obstacles are borne. Q

We can also improve the representation of this map by subjecting it to a morphological operation of the "closing" type (dilation - erosion), known in the field of image processing.

The card thus obtained thus provides precise information on the environment of the vehicle and allows:

- its evolution with possible avoidance of obstacles;

- its relocation by comparison between the measured locations of the obstacles and a priori knowledge of these obstacles.

It is also possible to envisage other signal transceiver systems than radar, for example laser or infrared rangefinders or sonars.

Claims

1. Method for detecting and locating obstacles in the environment of a vehicle, characterized in that it comprises the steps consisting in transmitting signals towards said environment from a plurality of radars (2) arranged on the vehicle, to deduce from the signals received in echo the probabilities of occupation by an obstacle of a set of elementary cells (9) of said environment, and to merge the probabilities relating to each of said cells.

2. Method according to claim 1, in which the fusion is carried out on the values of the probabilities provided by a plurality of radars, and in particular two radars.

3. Method according to any one of claims 1 and 2, in which the fusion is carried out on the value of the probability supplied by a radar and the value of the probability previously known by the same cell.

4. Method according to any one of claims 1 to 3, in which we deduce, from the echo signals received, distance information and angular information about the obstacles, and said occupation probabilities are determined from this information. .

5. Method according to any one of claims 1 to 4, comprising a step of thresholding the results of the merger.

6. Method according to any one of claims 1 to 5, in which the signal emitted by the radars is frequency modulated, the signal received in echo is mixed with a fraction of the signal emitted in order, after elimination of the carrier frequency, to form a beat signal, and the distance from an obstacle is obtained either by Fourrier transformation or by counting during a predetermined time interval.

7. Method according to any one of claims 1 to 6, wherein the emission of each radar lasts a limited time, and the repetition logic is coded for the identification of the source.

8. Device for detecting and locating obstacles in the environment of a vehicle, characterized in that it comprises a plurality of radars (2) mounted on the vehicle and a processing unit (7) arranged for merge data provided by speed cameras.

9. Device according to claim 8, in which the radars are wide-lobe.