WO2024032278A1

WO2024032278A1 - Laser radar, resource allocation method for laser radar, and computer readable storage medium

Info

Publication number: WO2024032278A1
Application number: PCT/CN2023/105700
Authority: WO
Inventors: 陈杰; 向少卿
Original assignee: 上海禾赛科技有限公司
Priority date: 2022-08-10
Filing date: 2023-07-04
Publication date: 2024-02-15
Also published as: CN117630866A

Abstract

A laser radar (10), a resource allocation method for the laser radar (10), and a computer readable storage medium. The laser radar (10) comprises: a plurality of transmitting units (11), configured to transmit detection beams; a plurality of detection units (12, 12A^~, 12A, 12B, 12C, 12D, 12E), which form a plurality of detection channels together with the plurality of transmitting units (11), the plurality of detection units (12, 12A^~, 12A, 12B, 12C, 12D, 12E) being configured to receive echoes of the detection beams transmitted by the transmitting units (11) and reflected by an obstacle; and a plurality of processing units (13, 13A, 13B, 13C, 13D, 13E), connected to the detection units (12, 12A^~, 12A, 12B, 12C, 12D, 12E), and configured to process the echoes received by the detection units (12, 12A^~, 12A, 12B, 12C, 12D, 12E) and generate a point cloud, wherein an echo processing relationship between the detection units (12, 12A^~, 12A, 12B, 12C, 12D, 12E) and the processing units (13, 13A, 13B, 13C, 13D, 13E) can be dynamically adjusted. The laser radar (10) comprehensively considers hardware overhead and algorithm implementation, uses as few measurement resources as possible to process changes in resolution and point cloud data rate brought by dynamically tracking one or more regions, and is particularly suitable for the requirements of engineering for balance among cost, volume and efficiency.

Description

LiDAR, LiDAR resource allocation method and computer-readable storage medium

Technical field

The present disclosure relates to the field of photoelectric detection, and in particular to a laser radar, a resource allocation method of the laser radar, and a computer-readable storage medium.

Background technique

Lidar is a commonly used ranging sensor with the advantages of long detection range, high resolution, strong anti-active interference ability, small size, and light weight. It is widely used in fields such as intelligent robots, drones, and autonomous driving. As a three-dimensional measurement system, lidar achieves three-dimensional measurement coverage of the measured field of view (FOV, Field of View) through collected point clouds. Multi-channel lidar based on Time of Flight (ToF) is suitable for situations where scanning a large field of view and acquiring high-density point clouds are required.

In radars with multiple channels working in parallel, in order to save TOF measurement resources, a commonly used solution is: multiple non-parallel channels share the same measurement resource in a time-shared manner, and channels working in parallel use different measurement resources. For clear explanation, Figure 1 shows a schematic diagram of the echo signal processing channel of the tree MUX architecture. Channels using the same measurement or processing resources are generally defined as a group. Each channel in the group works in time sharing. A channel in a different group Or multiple channels may work in parallel. For example, a group includes channels CH1-16, which work in a time-sharing and staggered manner and share the same measurement resource. Select a tree-shaped multiplexer (MUX, Multiplex) through address mapping to allow the 16 channels in the BANK to work in sequence. Timing, sequentially access signal processing channels, that is, measurement resources. The basic characteristics of the tree MUX are: the set of transceiver channels covered by each echo processing resource do not cross each other, and the connection relationship is fixed. .

However, there are some situations where the echo processing resources at the receiving end need to be dynamically configured. For example, in the field of autonomous driving, the field of view of lidar will be further divided into areas that require high attention or regions of interest (ROI, Region of Interests), such as target areas along the driving path. Moreover, if the ROI area can be moved quickly and dynamically or needs to be dynamically anchored and then scanned more precisely, it means that the number of echoes that need to be processed in the ROI area The data volume will be different when the ROI is not selected. At this time, the echo processing resources of the corresponding receiving end need to be dynamically configured. Or, if the horizontal or vertical scanning parameters of the lidar, such as the scanning frequency, are adjusted, adaptive allocation of echo processing resources will also be required.

However, if you want to allocate dynamic TOF measurement resources, the technical problem you will face is: because the engineering needs to balance cost, volume and efficiency, when some channels do not correspond to the ROI area, the echo output by the detector will If the amount of data is relatively small, there will be a minimum value; but when some channels correspond to the ROI area, the amount of echo data generated increases significantly, and there will be a maximum value. If the radar machine is configured with measurement resources for each channel according to the maximum value, the efficiency will be high, but the cost and volume will increase; if the measurement resources are configured for each channel according to the minimum value, the cost and volume will be low, but efficiency will be sacrificed.

The content in the background art section only discloses technologies known to the inventor, and does not necessarily represent the prior art in the field.

Contents of the invention

When dynamically tracking one or more areas, the echo and point cloud data volume of the area to be tracked is very different from that of other areas. Especially in engineering, it is necessary to strike a balance between cost, volume and efficiency, and to solve the problem of how to real-time Reasonably allocate measurement resources to cope with changes in resolution and point cloud data rate in different areas. Therefore, the present invention provides a lidar, including:

a plurality of transmitting units configured to transmit detection beams;

A plurality of detection units configured to receive echoes of detection beams emitted by the transmitting unit reflected by obstacles; and

A plurality of processing units, connected to the detection unit, configured to process the echoes received by the detection unit and generate point clouds;

Among them, the echo processing relationship between the detection unit and the processing unit can be dynamically adjusted.

According to one aspect of the present invention, at least one of the detection units is connected to multiple processing units, and for a detection unit connected to multiple processing units, its echoes can be shared and processed by multiple processing units.

According to an aspect of the present invention, when the field of view of the lidar includes a first area and a second area, and the amount of echo data in the first area is greater than the amount of echo data in the second area, the At least one detection unit in the first area is connected to a plurality of processing units.

According to an aspect of the present invention, the first area may include multiple sub-areas, and at least one detection unit corresponding to each sub-area is connected to multiple processing units.

According to an aspect of the invention, the plurality of sub-regions do not overlap with each other.

According to an aspect of the present invention, each detection unit corresponding to the first area is connected to a plurality of processing units.

According to an aspect of the present invention, each detection unit corresponding to the second area is connected to a processing unit.

According to one aspect of the invention, each processing unit is connected to a plurality of detection units.

According to an aspect of the present invention, the lidar further includes a plurality of multiplexing units, the plurality of detection channels are divided into multiple groups, and the detection units of each group are connected to a processing unit through a multiplexing unit. , multiple groups of detection units are connected to one processing unit through a multiplexing unit.

According to one aspect of the present invention, according to the expected point cloud results, multiple scanning modes are preset correspondingly, and the minimum number of processing units that need to be connected to the detection unit that satisfies each scanning mode is determined.

According to one aspect of the present invention, the processing unit connected to each detection unit is determined based on the echo data processing duration of each detection unit.

According to one aspect of the present invention, the processing unit connected to each detection unit is determined according to the rotation speed and point cloud resolution of the lidar.

According to one aspect of the present invention, the processing unit connected to each detection unit is determined according to the physical distance between the detection unit and the processing unit.

The present invention also provides a resource allocation method for a laser radar. The laser radar includes a plurality of transmitting units, a plurality of detection units and a plurality of processing units. The plurality of transmitting units are configured to transmit detection beams. The resource allocation method include:

The plurality of detection units receive echoes of the detection beams sent by the transmitting unit reflected by obstacles;

assigning the echo to one or more processing units connected to the detection unit to process the echo and generate a point cloud;

According to an aspect of the present invention, the resource allocation method further includes: when the field of view of the lidar includes a first area and a second area, and the amount of echo data in the first area is greater than the second area When the amount of echo data is equal to the amount of echo data, the echoes in the first area are allocated to one or more processing units connected to the detection unit corresponding to the first area.

According to an aspect of the present invention, the first area may include multiple sub-areas, and the resource allocation method further includes: allocating the echo data in the multiple sub-areas according to the amount of echo data in the multiple sub-areas. One or more processing units connected to detection units corresponding to the plurality of sub-areas.

According to an aspect of the present invention, the resource allocation method further includes: allocating the echo data in the second area to the detection unit connections corresponding to the second area according to the amount of echo data in the second area. processing unit.

According to an aspect of the present invention, the plurality of detection channels are divided into multiple groups, and the resource allocation method further includes: according to the amount of echo data of each group of detection channels, allocate the echo data of the group of detection channels to the corresponding group. A group detection channel connects one or more processing units.

According to an aspect of the present invention, the resource allocation method further includes: presetting multiple mapping tables according to point cloud distribution, the mapping table including the range of the first area and the mapping table according to the amount of echo data in the first area. The echo data in the first area is distributed to the processing unit connected to the detection unit corresponding to the first area.

According to one aspect of the present invention, the resource allocation method further includes: switching matching mapping tables according to changes in point cloud distribution.

According to an aspect of the present invention, the resource allocation method further includes: switching the mapping table between two probes.

The present invention also provides a computer-readable storage medium, including computer-executable instructions stored thereon. The executable instructions implement the resource allocation method as described above when executed by a processor.

This invention comprehensively considers hardware overhead and algorithm implementation, and uses as few measurement resources as possible to process the changes in resolution and point cloud data rate caused by dynamic tracking of one or more areas. It is especially suitable for engineering projects that balance cost, volume and efficiency. Balance needs.

Description of drawings

The accompanying drawings that form a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure. In the attached picture:

Figure 1 shows a schematic diagram of the echo signal processing channel of the tree MUX architecture;

Figure 2a shows a schematic diagram of the vertical field of view of lidar according to an embodiment of the present invention;

Figure 2b shows a schematic diagram of the laser radar horizontal field of view according to one embodiment of the present invention;

Figures 3a and 3b show a schematic diagram of a lidar according to an embodiment of the present invention;

Figure 3c shows a schematic diagram of a lidar according to another embodiment of the present invention;

Figures 4a-4c show schematic diagrams of the connection relationship between the detection unit and the processing unit according to multiple embodiments of the present invention;

Figure 5a shows a schematic diagram of the division of the first area and the second area according to an embodiment of the present invention;

Figure 5b shows a schematic diagram of a first area including multiple sub-areas according to an embodiment of the present invention;

Figure 5c shows a schematic diagram of a first area including a plurality of separated sub-areas according to an embodiment of the present invention;

Figure 5d shows a schematic diagram of a first area including a plurality of separated sub-areas according to another embodiment of the present invention;

Figures 6a-6c show a schematic diagram of a scanning mode according to an embodiment of the present invention;

Figure 7 shows a schematic diagram of a scanning mode according to another embodiment of the present invention;

Figure 8 shows a schematic diagram of first area wandering according to an embodiment of the present invention;

Figures 9a-9c show a schematic diagram of the connection relationship between the BANK and the processing unit of Figure 9 corresponding to multiple embodiments of the present invention;

Figure 10a shows the echo data under the connection of scanning mode index = 1 according to an embodiment of the present invention. Allocation diagram;

Figure 10b shows a schematic diagram of echo data distribution under a connection with scanning mode index=8 according to an embodiment of the present invention.

Detailed ways

In the following, only certain exemplary embodiments are briefly described. As those skilled in the art would realize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the invention. Accordingly, the drawings and description are to be regarded as illustrative in nature and not restrictive.

In the description of the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", " The directions indicated by "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" etc. or The positional relationship is based on the orientation or positional relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, Therefore, it should not be construed as a limitation of the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, features defined as "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "plurality" means two or more than two, unless otherwise clearly and specifically limited.

In the description of the present invention, it should be noted that, unless otherwise clearly stated and limited, the terms "installation", "connection" and "connection" should be understood in a broad sense. For example, it can be a fixed connection or a detachable connection. Connection, or integral connection: it can be mechanical connection, electrical connection or mutual communication; it can be directly connected, or it can be indirectly connected through an intermediary, it can be the internal connection of two elements or the interaction of two elements relation. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

In the present invention, unless otherwise clearly stated and limited, a first feature "on" or "below" a second feature may include the first and second features being in direct contact, or may include the first and second features. No Not in direct contact but through other characteristic contacts between them. Furthermore, the terms "above", "above" and "above" the first feature "above" the second feature include the first feature being directly above and diagonally above the second feature, or simply means that the first feature is higher in level than the second feature. "Below", "below" and "beneath" the first feature of the second feature includes the first feature being directly above and diagonally above the second feature, or simply means that the first feature is less horizontally than the second feature.

The following disclosure provides many different embodiments or examples of various structures for implementing the invention. In order to simplify the disclosure of the present invention, the components and arrangements of specific examples are described below. Of course, they are merely examples and are not intended to limit the invention. Furthermore, the present invention may repeat reference numbers and/or reference letters in different examples, such repetition being for purposes of simplicity and clarity and does not itself indicate a relationship between the various embodiments and/or arrangements discussed. In addition, the present invention provides examples of various specific processes and materials, but one of ordinary skill in the art will recognize the application of other processes and/or the use of other materials.

Figure 2a shows a schematic diagram of the vertical field of view of a lidar according to an embodiment of the present invention. Multiple lasers of the lidar emit light according to a certain timing sequence. After the emitted light is collimated and processed by the emitting lens (group), from When the lidar emits, it points in different directions. In a lidar where a laser and a detector form a detection channel, each channel is responsible for scanning a certain vertical angular range. The vertical field of view is the total angular range that the lidar can detect in the vertical direction. All channels correspond to The sum of the vertical angle ranges together constitutes the vertical field of view of the radar. In Figure 2a, n channels/harnesses are shown, n can be 16 or 32 or 40 or 64 or 128 or more or less, and the vertical field of view is (x+y)°, covering -x°~+ y°, consisting of the vertical angles of all lasers combined. Among them, the vertical direction angle of the top laser is +y° (+ means upward relative to the horizontal plane), which is responsible for distance detection in the y° direction, and the vertical direction angle of the bottom laser is -x°, which is responsible for Distance detection in x° direction downward. In specific implementation, the actual vertical field of view range of lidar can be 40°, 100°, or other values.

Figure 2b shows a schematic diagram of the horizontal field of view of the lidar according to an embodiment of the present invention. The figure is a cross-sectional view of the lidar in the horizontal direction. The horizontal direction is generally perpendicular to the rotation axis of the radar. The radar Driven by rotating components such as motors, or through rotating mirrors, galvanometers, MEMS, or liquid crystals, the entire horizontal field of view can be scanned. The horizontal field of view is the angular range that the lidar can detect in the horizontal direction. For example, if a mechanical lidar rotates 360°, the horizontal field of view is 360°. In the point cloud image output by the lidar, the angle between two adjacent detection points on the horizontal plane perpendicular to the rotation axis is the horizontal angular resolution. In Figure 2b, the cross-section of the radar is circular. The present invention is not limited thereto. The cross-section of the radar can also be in other shapes, such as rectangular. Moreover, the rotating axis (or rotating component) can be arranged at the center of the radar, or at a position to the left or to the right, which are all within the scope of the present invention.

The vertical field of view and the horizontal field of view together constitute the field of view of the lidar. The detection method of the present invention is not only suitable for mechanical lidar, but also for solid-state lidar and semi-solid state lidar. Taking solid-state lidar as an example, its horizontal field of view is usually less than 360 degrees, such as 120 degrees, and includes multiple lasers and multiple detectors, forming multiple detection channels.

The present invention provides a laser radar, which includes: a plurality of transmitting units configured to transmit detection beams; a plurality of detection units forming multiple detection channels with the plurality of transmitting units; the plurality of detection units are configured to receive detection signals sent by the transmitting units. The echo of the light beam reflected by the obstacle; and a plurality of processing units, connected to the detection unit, configured to process the echo received by the detection unit and generate a point cloud; wherein, the echo processing of the detection unit and the processing unit Relationships can be adjusted dynamically. At least one detection unit is connected to multiple processing units, and for a detection unit connected to multiple processing units, its echo data can be shared and processed by multiple processing units. This invention comprehensively considers hardware overhead and algorithm implementation, and uses as few measurement resources as possible to process the changes in resolution and point cloud data rate caused by dynamic tracking of one or more areas. It is especially suitable for engineering projects that balance cost, volume and efficiency. Balance needs.

The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described here are only used to illustrate and explain the present invention, and are not intended to limit the present invention.

3a and 3b show a schematic diagram of a lidar according to an embodiment of the present invention. The lidar 10 includes a plurality of transmitting units 11, a plurality of detection units 12 and a plurality of processing units 13 (not shown in Fig. 3a, located in The back of the RX board, arranged opposite to the detection unit), the details are as follows:

The plurality of transmitting units 11 are configured to transmit detection beams, and the plurality of detection units 12 are connected with the plurality of transmitting units. The unit 11 constitutes multiple detection channels, each detection channel corresponds to a vertical orientation, and the multiple detection units 12 are configured to receive the echo data of the detection beam reflected by the obstacle from the transmitting unit 11 located in the same detection channel. For example, multiple detection units 12 and multiple transmitting units 11 form multiple detection channels, and each detection channel includes at least one transmitting unit 11 and at least one detection unit 12 . Multiple detection channels can emit light in parallel, or be divided into multiple group BANKs. The emission units 11 in each BANK emit light in non-parallel, and only one emission unit 11 in each BANK emits light at the same time. The lighting sequence of the emission unit 11 can also be controlled through timing, which are all within the protection scope of the present invention.

A plurality of processing units 13 are connected to the detection unit 12 and configured to process the echo data and generate point clouds.

Among them, the emission unit includes a laser, and the specific type can be VCSEL or EEL. The detection unit includes an optoelectronic device to convert the received optical signal into an electrical signal. For example, APD, SiPM or Spad array can be used. The processing resources can perform subsequent processing on the electrical signal output by the optoelectronic device, such as analyzing and obtaining the time of arrival of the echo and the power of the echo, etc., and then obtaining the distance and/or reflectivity information of the obstacle; or for the transmitting end. In other words, the processing resource may be a driving circuit that drives the laser to emit light. Depending on the type of optoelectronic device used, different processing units can be selected. For example, for devices such as APD and SiPM that output analog signals, ADC can be used for further processing of subsequent signals; while for devices such as SPAD arrays that output digital signals, TDC can be used as the processing unit. In addition, the processing unit in this article is a device that has an upper limit on the amount of echo data/laser drive amount within a certain period of time. That is, within a certain period of time, the processing unit has limited processing resources, and the amount of echoes that can be processed per unit time/ Drivers are processing resources. Over a period of time, the processing unit can allocate its total processing resources to drive different lasers or process echo data from different detectors. The upper limit of processing resources of each processing unit can be the same or different. For the convenience of description, the following description in this article will assume that the number of processing units is positively related to the number of processing resources, that is, allocating more processing units equals allocating more processing resources. In addition, it should be noted that regarding processing resources, this article takes the receiving end as an example. The actual transmitting end is the same and will not be repeated.

Among them, in order to support scenarios where lidar needs to dynamically adjust scanning parameters, the detection unit and processing The echo processing relationship of the management unit can be set to be dynamically adjustable. Specifically, at least one detection unit 12 can be connected to multiple processing units 13 , and then for the detection unit 12 connected to multiple processing units 13 , the data obtained after detecting the echo can be processed by multiple processing units 13 Processing is shared, and data processing tasks can be dynamically assigned to one or more of the plurality of processing units 13 for execution, thereby meeting the requirements of real-time data processing and data processing load requirements.

In a specific embodiment, referring to Figure 4a, the lidar 10 includes four detection units 12, namely the detection unit 12A, the detection unit 12B, the detection unit 12C and the detection unit 12D. The lidar 10 also includes three processing units 13, namely a processing unit 13A, a processing unit 13B and a processing unit 13C. For example, due to the requirement of high detection resolution, the detection unit 12A has a very large amount of echo data within a certain period of time (for example, for a region of interest ROI, the detection unit 12 and the corresponding laser are responsible for the region of interest. ROI is scanned more intensively, resulting in a larger amount of echo data), and within this period of time, these point cloud data cannot be processed in time by relying solely on the processing unit 13A or 13B, and because the processing unit 13A may still Echoes from other detectors 12A~ need to be processed, and only part of the resources can be used to process echoes from the detection unit 12A. Therefore, in the embodiment of Figure 4a, the detection unit 12A is connected to the processing unit 13A and the processing unit 13B, then the echo data received by the detection unit 12A can be shared and processed by the two processing units 13A and 13B to ensure that all current The processing unit, each processing unit is not overloaded, and the data of all detectors in the laser radar such as 12A~, 12A, 12B, 12, 12D and other not shown can be processed within this period of time. Furthermore, the amount of echo data of the detection unit 12B, the detection unit 12C and the detection unit 12D is often relatively small, and the total amount of point cloud data can be borne by one processing unit 13, then the three are connected to the processing unit 13C respectively. The echo data received are independently processed by the processing unit 13. As can be seen from Figure 4a, four detection units 12 are connected to three processing units 13. Compared with the solution in which one detection unit 12 is connected to one processing unit 13, one processing unit 13 is saved, and TOF measurement resources are reasonably allocated in real time. Taking over all the echo data; compared with the solution of one processing unit 13 connecting multiple detection units 12, it avoids the problem that the processing unit 13 cannot process the echo data in time, and achieves efficient allocation of TOF measurement resources.

In addition, in the embodiment of Figure 4a, the detection unit 12A and the processing unit 13A and the processing unit 13B is connected, the echo generated by the detection unit 12A can be processed by the processing unit 13A alone, or can be processed by the processing unit 13B alone, or can also be shared by the processing unit 13A and the processing unit 13B at the same time (for example, according to a certain ratio) . The allocation of specific data processing tasks may depend on the amount of echo data that needs to be processed and the current task load of each processing unit 13A and 13B. For example, when the amount of echo data currently generated by the detection unit 12A is small (less than a preset value), the echo data generated by the detection unit 12A can be processed individually by the processing unit 13A or the processing unit 13B (for example, according to a preset sequence selection); when the amount of echo data currently generated by the detection unit 12A is large (greater than the preset value), it can be dynamically allocated between the processing unit 13A and the processing unit 13B according to the preset ratio, thereby ensuring the efficiency of echo processing. real-time.

It should be noted that, for the transmitting end, the measurement resource is the laser drive circuit channel; for the receiving end, it is the echo processing resource, such as the ADC or TDC channel, that is, the channel that assists TOF calculation. In addition, the amount of echo data in the ROI area and the non-ROI area is very different. The horizontal resolution of the ROI area can be x times that of the non-ROI area, and the vertical resolution can be y times that of the non-ROI area. The overall point cloud data is xy times that of the non-ROI area. .

In another specific embodiment, referring to Fig. 4b, the lidar 10 includes four detection units 12, namely the detection unit 12A, the detection unit 12B, the detection unit 12C and the detection unit 12D. The lidar 10 also includes four processing units 13, namely a processing unit 13A, a processing unit 13B, a processing unit 13C and a processing unit 13D. For example, the amount of echo data of the detection unit 12A is often very large, and it is connected to the processing unit 13A and the processing unit 13B. Then the echo data received by the detection unit 12A can be shared and processed by the two processing units 13 . Furthermore, the amount of echo data of the detection unit 12B is also relatively large, but one processing unit 13 is enough to bear it, so the detection unit 12B is connected to the processing unit 13C and has exclusive use of this processing unit; the echoes of the detection unit 12C and the detection unit 12D The amount of data is often relatively small, and the total amount of point cloud data can be borne by one processing unit 13, then the two are respectively connected to the processing unit 13D, and the echo data received by the two are independently processed by the processing unit 13D. As can be seen from Figure 4b, four detection units 12 are connected to four processing units 13. Compared with the solution in which one detection unit 12 is connected to one processing unit 13 or one processing unit 13 is connected to multiple detection units 12, a reasonable use of measurement resources is achieved. and efficient configuration.

In another specific embodiment, referring to Figure 4c, the lidar 10 includes five detection units 12, namely, the detection unit 12A, the detection unit 12B, the detection unit 12C, the detection unit 12D and the detection unit 12E. The lidar 10 also includes five processing units 13, namely a processing unit 13A, a processing unit 13B, a processing unit 13C, a processing unit 13D and a processing unit 13E. For example, the amount of echo data of the detection unit 12A is often very large, and it is connected to the processing unit 13A and the processing unit 13B. Then the echo data received by the detection unit 12A can be dynamically shared and processed by the two processing units 13 . Furthermore, the amount of echo data of the detection unit 12B and the detection unit 12C is equivalent and relatively large. The detection unit 12B is connected to the processing units 13C and 13D, and the detection unit 12C is connected to the processing units 13C and 13D, which can realize a detection unit. The echo data of 12 is dynamically shared by two processing units 13, and one processing unit 13 can assist two detection units 12. The amount of echo data of the detection unit 12D and the detection unit 12E is often relatively small, and the total amount of point cloud data can be borne by one processing unit 13, then the two are connected to the processing unit 13E respectively, and the echo data received by the two is processed by the processing unit 13E. The processing unit 13 processes independently. As can be seen from Figure 4c, five detection units 12 are connected to five processing units 13. Compared with the solution in which one detection unit 12 is connected to one processing unit 13 or one processing unit 13 is connected to multiple detection units 12, TOF measurement resources are realized. Reasonable and efficient configuration.

The feature of "at least one of the detection units 12 is connected to multiple processing units 13, and for the detection unit 12 connected to multiple processing units 13, the echo data can be shared and processed by multiple processing units 13" is introduced above through the embodiment. , the common point of the above embodiments is that the detection unit 12A is connected to two processing units 13, and the echo data received by it is shared by the two processing units 13. Furthermore, it is allocated to other detection units 12 according to the amount of echo data. The processing unit 13 realizes efficient allocation of measurement resources and avoids the problem that one detection unit 12 is connected to one processing unit 13, some processing units 13 have insufficient resources, and some processing units 13 have excessive resources. Those skilled in the art can understand that the numbers in the above embodiments are only illustrative descriptions, the connection relationship between the detection unit and the processing unit and the allocation of processing resources according to the amount of echo data are illustrative descriptions, and do not constitute a limitation of the present invention. limits.

The above provides a macro introduction to the system architecture and resource allocation strategy of LiDAR 10. The amount of echo data for different areas varies greatly, and the area of interest needs to be quickly and dynamically tracked. The demand for more refined scanning requires optimization of measurement resources, especially the processing unit 13 at the receiving end needs to be dynamically configured. Therefore, the hardware connections between the plurality of processing units 13 and the plurality of detection units 12 and the dynamic distribution of echo data are the focus of the present invention. In the present invention, "processing unit" should be understood in a broad sense, including both hardware resources and data computing and processing capabilities that each hardware resource can provide in time sharing or in parallel. For example, the multiple processing units described in Figures 4a-4c may include multiple data processing devices, such as hardware devices such as TDC and/or ADC, or may include resources with data computing and processing capabilities provided by one data processing device. . The former is easy to understand, and the latter will be described below. For a data processing device, its data processing capability can be divided into different time slots, and different time slots can be allocated to different detection units 12, that is, used to process echoes generated by different detection units 12. data. In this case, one time slice of one data processing device may correspond to one processing unit shown in Figures 4a-4c, and multiple time slices of one or more data processing devices may correspond to the processing units shown in Figures 4a-4c. A plurality of processing units 13 are provided. In addition to time slices, the data processing device can also run different threads at the same time to process the echo data generated by different detection units 12 respectively. In this case, one thread of one data processing device corresponds to one processing unit shown in Figures 4a-4c, and multiple threads of one or more data processing devices corresponds to the processing units shown in Figures 4a-4c. Multiple processing units 13. In addition, the time slices or threads of the data processing equipment can also be synchronized, that is, they can be used to process echo data generated by different detection units 12 at the same time. Therefore, the "processing unit" in the present invention should be understood to have a broad meaning.

In order to effectively utilize the processing unit 13, the concept of resource pool is introduced, that is, all processing units 13 are planned as a total resource pool, specifically, the hardware and software processing capabilities (such as time slices and threads) of all processing units 13 are It is planned as a resource pool, in which multiple processing units 13 can be connected to the same detection unit 12, so that the multiple processing units 13 can share the echo data of the detection unit 12. The processing unit mentioned in this specification can refer to either a physical entity or an intangible processing resource, which does not limit the present invention.

Although the above embodiment exemplarily illustrates the connection relationship between the detection unit 12 and the processing unit 13, it can cope with the situation where the amount of echo data received by each detection unit 12 is known in advance and the amount of echo data is relatively fixed. If the amount of data is dynamically tracked and configured, further steps are required. Refine wiring and resource allocation plans. But before explaining this solution, we first introduce the point cloud pattern and regional division of the lidar field of view, and introduce the hardware connection relationship between the detection unit and the processing unit based on the regional division.

According to a preferred embodiment of the present invention, when the field of view of the lidar includes a first area and a second area, and the amount of echo data in the first area is greater than the amount of echo data in the second area, the corresponding At least one detection unit is connected to a plurality of processing units.

Figure 5a shows a schematic diagram of the division of the first area and the second area according to an embodiment of the present invention. The largest rectangular frame is the field of view of the laser radar 10. The horizontal field of view is the horizontal field of view, and the longitudinal field is the vertical field of view. Select the One area is the first area, and the other areas are the second area. The first area can be set according to the area of common concern or the area where the obstacles of interest are located when the radar is used in ADAS scenarios, such as directly in front of the vehicle. Obviously, the amount of echo data in the first area will be greater than the amount of echo data in the second area. More processing units 13 should be allocated to the detection units 12 corresponding to the first area to at least satisfy at least one detection unit corresponding to the first area. The unit is connected to multiple processing units.

However, in practice, the obstacles of real interest only occupy a small part of the area. If the first area is too large, too many processing units 13 may be allocated to it, and uneven distribution of processing resources will inevitably occur. question.

According to a preferred embodiment of the present invention, the first area may include multiple sub-areas, and at least one detection unit corresponding to each sub-area is connected to multiple processing units.

Figure 5b shows a schematic diagram of the first area including multiple sub-areas according to an embodiment of the present invention. The first area is further divided into multiple sub-areas, such as sub-areas 1-4. Only one or more sub-areas are activated for each detection. The so-called activated means that in the activated sub-area, due to the increase in the number of scanning line beams and/or the increase in scanning frequency, the amount of echo data (or the density of echo reception or the total number of echoes within a period of time) is relatively large. Inactive subregions are higher. For example, if an obstacle appears in sub-area 1, the amount of echo data in sub-area 1 is relatively high. If the user chooses to dynamically track the obstacle, sub-area 1 will be activated, that is, the scanning line beam in sub-area 1 will be activated. Or the scanning frequency will be increased to obtain a higher resolution point cloud for the user-selected obstacles, so the amount of echo data generated is relatively larger, and the amount of echo data that each processing unit can process per unit time There is an upper limit, so it needs to be a sub-region 1 The corresponding detection unit 12 is allocated more processing units 13 (that is, the echoes generated by the detection unit 12 are allocated to multiple processing units 13 connected to it for processing, or one of the processing units 13 is allocated for a period of time. Only the data of this detection unit 12 are processed exclusively). The amount of echo data in sub-regions 2-4 and the second region is relatively small, and each corresponding detection unit 12 requires fewer processing units (therefore, taking one detection unit 12 in sub-region 2-4 as an example) , even though it may be connected to multiple processing units, since the amount of echo data is small, its echo data can be assigned to only one of the connected processing units 13 for processing). It should be noted that if the point cloud data of a sub-region is high, it means that the horizontal or vertical resolution of the sub-region has been improved. If it is improved horizontally, it may be by increasing the transmission frequency, then the corresponding detector reception frequency It also needs to be improved, which means that the detector receives more echo data within a certain period of time; if it is improved vertically, it may be that a certain detector previously received an echo within the period of t1, and currently it is within the period of t2. After receiving one echo, t1>t2, the echo data received by the detector per unit time is also higher.

According to a preferred embodiment of the present invention, the plurality of sub-regions do not overlap with each other, and the plurality of sub-regions can be detected simultaneously.

Figure 5c shows a schematic diagram in which the first area includes multiple separate sub-areas according to an embodiment of the present invention. The difference from the embodiment in Figure 5b is that the first area is divided into multiple separate sub-areas. Figure 5c is only an exemplary illustration. This embodiment does not limit the size, coverage, number of sub-regions or the relative positions between sub-regions. For example, if an obstacle appears in sub-area 1 while the vehicle is driving, the amount of echo data in sub-area 1 is relatively high. If the user selects the area of interest to dynamically track the obstacle, activate sub-area 1. , that is, the scanning beam or scanning frequency in sub-area 1 will be increased to obtain a higher-resolution point cloud for the obstacles selected by the user, so the amount of echo data generated is relatively larger, and each processing unit The amount of echo data that can be processed per unit time is limited, so more processing units 13 need to be allocated to the detection unit 12 corresponding to sub-area 1 (that is, the echo generated by the detection unit 12 is allocated to multiple processing units connected to it) 13 for processing, or allocate one of the processing units 13 to exclusively process the data of the detection unit 12 in the next period of time). The amount of echo data in sub-area 4 and the second area is relatively small, and each corresponding detection unit 12 requires fewer processing units.

Figure 5d shows a schematic diagram of another first area of the present invention including multiple separated sub-areas. The difference from the embodiment of Figure 5c is that the first area is divided into multiple separate sub-areas with different sizes. . Figure 5d is only an exemplary illustration. This embodiment does not limit the size, coverage, number of sub-regions or the relative positions between sub-regions. For example, when the vehicle is driving and an obstacle appears in sub-area 4, the amount of echo data in sub-area 4 is relatively high. If the user selects the area of interest to dynamically track the obstacle, activate sub-area 4. , that is, the scanning beam or scanning frequency in sub-area 4 will be increased to obtain a higher-resolution point cloud for the obstacles selected by the user, so the amount of echo data generated is relatively larger, and each processing unit The amount of echo data that can be processed per unit time is limited, so more processing units 13 need to be allocated to the detection unit 12 corresponding to sub-area 4 (that is, the echo generated by the detection unit 12 is allocated to multiple processing units connected to it) 13 for processing, or allocate one of the processing units 13 to exclusively process the data of the detection unit 12 in the next period of time). The amount of echo data in sub-area 1 and the second area during this period is relatively small, and each corresponding detection unit 12 requires fewer processing units.

The above embodiments have introduced the division of the first area into multiple sub-areas, which can save computing power in locating the area to be tracked based on the selected area of interest and reduce system power consumption. However, when switching between multiple sub-areas, multiple scanning modes can also be preset to further save computing power and improve efficiency. Further introduction below.

Synchronous detection means that the lidar can complete the detection of multiple non-overlapping sub-areas in one detection process without switching or changing the field of view.

Some galvanometer lidars claim to be able to track multiple areas. The specific operation method is: reduce the field of view FOV of the galvanometer to match the size of the sub-area corresponding to the obstacle, such as the galvanometer itself. The FOV is 25° horizontally + 30° vertically. In order to scan this sub-area, reduce the FOV to 5° horizontally + 6° vertically, and then scan only within the smaller FOV. The result is: 1. Each scan can only focus on one obstacle, and only scan the sub-area corresponding to the obstacle. If there are two or more obstacles, multiple scans need to be performed; 2. While the sub-area corresponding to the obstacle is scanned, other areas will not be scanned because the FOV has been reduced to exclude other areas from the scanning range. Therefore, the difference between the present invention and the galvanometer laser radar is at least that: it can Set up multiple sub-areas that do not overlap each other, and can detect multiple sub-areas simultaneously in parallel. The definition of synchronization is that there is no need to switch the field of view range, so as to avoid scanning one sub-area and then switching to aiming at another sub-area. Then the sub-area is scanned. If the objects that appear within the time difference between the two scans move, accurate detection results cannot be obtained.

The above embodiment achieves further refinement of the first region by setting sub-regions, that is, refining the region that requires more processing units and the region that requires fewer processing units, thereby more accurately matching the user-selected sense. Obstacles or areas of interest. But in practice, the obstacle of interest is moving, or moves relative to the lidar. In order to track the obstacle of interest, the first area that moves within the lidar field of view can be set, and the first area The size is adjustable to take into account regional refinement and dynamic configuration of measurement resources.

According to a preferred embodiment of the present invention, multiple scanning modes are preset according to one or more of the number, size, moving direction and moving step of the sub-regions in the first area, and different scanning modes correspondingly activate different sub-regions. area.

Figures 6a-6c show a schematic diagram of a scanning mode according to an embodiment of the present invention. The largest rectangular frame represents the full field of view that can be achieved in a single scan. The horizontal field of view is the horizontal field of view and the vertical field of view is the vertical field of view. The dotted box in it Inside is the first area, and more processing units 13 should be allocated to the detection unit 12 corresponding to this area; the relatively smaller solid line rectangular frame is a sub-area that is activated once, that is, in one scan, it can actually be Once an activated sub-region is activated, more processing resources are allocated to the detection unit 12 corresponding to the region. This processing resource can be more processing units 13, or it can be allocated to one or more processing units 13 within a period of time. The multiple processing units 13 are used to only process the echo data from the detection unit; the area between the largest rectangular frame and the dotted rectangular frame is the second area. The amount of echo data in this area is small and can be the corresponding detection area. Unit 12 is allocated fewer processing units 13.

For example, in the advanced driving assistance system ADAS application scenario, the area directly ahead is of most concern, such as the first area shown in the dotted rectangular box. This embodiment further subdivides this area into multiple sub-areas that can be activated, as shown in Figure 6a A single activated sub-region 1 is shown, a single activated sub-region 2 is shown in Figure 6b and a single activated sub-region 3 is shown in Figure 6c. Preferably, they are arranged in sequence from the top to the bottom of the vertical field of view according to the direction of movement. For example, when the vehicle is driving, the selected area of interest is related to the vehicle As the distance between vehicles gets closer and closer, the scanning pattern can be matched, and the selected area of interest partially overlaps with the single-activated sub-areas 1-3 in turn. The greater the number of sub-areas that are activated at a time, the smaller the field of view covered by each sub-area, the smaller the movement step, and the more accurately it can match the selected area of interest. Among them, the moving direction of the sub-regions can be set according to practical data, and the size, coverage, moving direction and moving step length of multiple sub-regions can be different.

Figure 7 shows a schematic diagram of the scanning mode of another embodiment of the present invention. As a variant embodiment, there is only one activatable area, and the activatable area can move within the field of view of the lidar, as shown in the direction of the arrow. It can move horizontally, vertically, or in any direction. Therefore, multiple scanning modes are preset according to one or more of the number, size, moving direction and moving step of the activatable area, which can be freely configured in engineering. The difference from the previous embodiment is that: from the matching sub- Change to matching scan mode, which simplifies the control sequence and is more efficient in engineering applications.

According to a preferred embodiment of the present invention, a matching scanning mode is activated based on changes in the selected region of interest.

Continuing to refer to Figure 7, for example, one can define an index number for the scan mode, a number that activates all possible options for area movement. For example, there can be nine scanning modes. The field of view covered by the activateable area in each scanning mode is different, and they are named index 0-8 respectively. Then you only need to negotiate the parameters of each mode with the user, and the user can determine the object of interest or select the area of interest as needed, and then specify an index i (i∈0-8) to control the activateable area of movement.

In some embodiments, the lidar can be switched among 9 indexes at will. For example, index 0 is selected in the first round of scanning. After obtaining the point cloud, the area of interest selected based on the point cloud is located in the area covered by index 8. Near the field of view (partially overlapping with the field of view), that is, if the index 8 mode is used to maximize the coverage of the area of interest, then in the second round of scanning, you can switch to index 8. The switching performed in this embodiment is a switching of moving steps, and the first round of scanning and the second round of scanning both refer to completing a detection of the first field of view range.

The above embodiment introduces the area division and scanning mode within the lidar field of view, determines the distribution of point clouds, or in other words, determines the amount of echo data in different areas, thereby providing corresponding data for different areas. The detection unit 12 is dynamically allocated with more or less processing units 13 . The advantage of area division or scanning mode is that it takes into account both dynamics and cost, because it is difficult to achieve complete dynamics in engineering.

Faced with such area division or scanning mode, in order to use as few measurement or processing resources as possible to cope with the dynamically changing amount of echo data, it is the basis of the present invention to implement cross-connection between the detection unit 12 and the processing unit 13 in hardware. Further introduction below.

The amount of echo and point cloud data in the first region is greater than the amount of echo and point cloud data in the second region. In a specific embodiment, the amount of echo and point cloud data in the first region occupies the total echo and point cloud data. The amount of data is about 3/4. If a conventional tree MUX architecture solution is used to connect multiple detection units 12 and multiple processing units 13, it will be found that the processing unit corresponding to the first area has no time to process the echo data, and the processing unit in the second area has no time to process the echo data. Relatively idle, that is, there will be a problem of uneven allocation of echo processing resources. Therefore, in order to reasonably allocate echo processing resources in real time, the concept of resource pool is introduced, that is, all echo processing resources are planned as a total resource pool, that is, the connection between the processing unit 13 and the detection unit 12 can Crossover, that is, two or more processing units 13 can be connected to the same detection unit 12. According to the size of each area, its echo and the amount of point cloud data, multiple processing units 13 can share the load with each other.

According to a preferred embodiment of the present invention, each detection unit 12 corresponding to the first area is connected to a plurality of processing units 13.

According to a preferred embodiment of the present invention, each detection unit 12 corresponding to the second area is connected to a processing unit 13. The amount of data in the second area is very small compared to the amount of data in the first area and there is no dynamic control. There will be no situation where the processing unit 13 connected to it cannot process the echo data. Therefore, each detection unit in the second area 121 can be connected to a processing unit 13.

According to a preferred embodiment of the present invention, each processing unit is connected to multiple detection units.

The above three preferred embodiments will be further described below in conjunction with the accompanying drawings.

Figure 8 shows a schematic diagram of the first area wandering according to an embodiment of the present invention. This radar has a total of The second area includes (xn) lines (there are xn/2 lines in the upper area and lower area of the first area's wandering range). Divide the x-line into multiple group BANKs, each BANK includes multiple detection channels, according to The number of BANK is further divided into odd-numbered BANK and even-numbered BANK. The difference in vertical angle between each two lines is the vertical angular resolution. Taking even-numbered BANK as an example, even-numbered BANK includes x/2 lines, and even-numbered BANK travels in the first area. The range includes n/2 lines, and the upper and lower areas are (xn)/4 lines respectively.

The horizontal and vertical resolutions of the first area are both y times that of other areas except the first area, and the total is y ² times the point cloud data. The first area can move within the range of n lines × vertical angular resolution (for example, 0.1°) of the vertical field of view. The current width of the first area is a line × 0.1°/line, and the moving step is b line/ times, walking from the bottom to the top, the number of steps required = (n line - line a)/b = c steps, so walking in the entire dotted line area corresponds to (c+1) modes (current mode The first area is at the bottom of the dotted area).

Figures 9a-9c show a schematic diagram of the connection relationship between BANK and the processing unit corresponding to Figure 8 according to multiple embodiments of the present invention. Generally speaking, the number of lines x of the radar is determined, and what needs to be optimized is the number of processing units 13 And the connection relationship between the processing unit 13 and the detection unit 12. Specifically, the x-line is divided into multiple group BANKs, each BANK includes multiple detection channels, and can be further divided into odd-numbered BANKs and even-numbered BANKs according to the number of the BANK.

In Figure 9a, taking the even-numbered BANK as an example (the odd-numbered BANK is symmetrically designed), there are a total of 8 BANKs, namely BANK 0-14, among which BANK 0 and BANK 14 correspond to the second area, and BANK 2-12 correspond to the games in the first area. According to the scanning mode, it is determined how to distribute the echo data of the detectors of which banks to the processing unit 13 connected thereto for processing. This embodiment does not limit the number of detection units 12 and transmitting units 11 in each BANK. As an illustrative embodiment, for example, BANK 2 includes 8 detection units 12 and 8 corresponding transmitting units 11. BANK 8 includes 12 detection units 12 and corresponding 12 transmitting units 11. The transmitting units 11 in the same BANK work in time sharing. For example, one transmitting unit 11 is selected from each BANK at a time, and the transmitting units 11 from other BANKs are The emission units 11 emit light in parallel.

According to a preferred embodiment of the present invention, as shown in Figure 3c, the lidar 10 also includes a plurality of multiplexing units 14. The plurality of detection channels are divided into a plurality of groups BANK, and the detection units 12 of each group BANK A processing unit 13 is connected via a multiplexing unit 14 , and the detection units 12 of a plurality of groups of BANKs are connected to a processing unit 13 via a multiplexing unit 14 .

The technical concept of the present invention is to form a cross-interconnected network between the BANK and the processing unit 13 . The so-called cross-connection is different from the fixed connection through the tree MUX shown in Figure 2, but means that each BANK can be connected to multiple processing units 13, and any one processing unit 13 can also be connected to multiple BANKs. . In other words, a cross MUX network is used to connect multiple processing units 13 and multiple BANKs. The so-called cross means that at least one processing unit 13 is connected to multiple BANKs, and at least one BANK is connected to multiple processing units 13 . In this way, the connection relationship between BANK and the processing unit 13 is relatively dynamic and not fixed. Reference can be made to Figures 9a-9c.

In addition, each BANK can include multiple detection channels, and a tree-shaped MUX can be selected between multiple detection channels. That is to say, each specific detection channel is patrolled through the MUX, and multiple detection channels are time-shared. Work, take turns sending and receiving. However, the present invention does not limit the number of BANKs and the number of detection channels in each BANK, nor does it limit the basis and method of area division.

According to a preferred embodiment of the present invention, according to the expected point cloud results, corresponding to multiple preset scanning modes, the minimum number of processing units 13 that need to be connected to the detection unit 12 that satisfies each scanning mode is determined.

Continuing to refer to Figure 9a, the lidar includes four processing units, namely processing units 13A-13D. In a certain detection, the order of processing resource allocation can cooperate with the scanning mode of the first area. For example, BANK 6 is connected to 13A, 13B, and 13C. It may be that in a certain detection, the total number of points generated by BANK 6 is 3 If there are relatively few points, all the echo data of BANK 6 will be allocated to 13A for processing; or in the next detection, the total number of points generated by BANK 6 will be 20, which is relatively large, so the echo data of BANK 6 will be allocated to 13A for processing. Part of the data, such as 8, is allocated to 13B for processing, and another part, such as 12, is allocated to 13C for processing. It should be noted that a detection process, that is, one round of transmission and one round of reception, will take a certain amount of time. Within this time, there is an upper limit to the amount of echo data that each processor can process, but each processor can The amount of echo data processed can vary. The total amount of echo data that all processors can process within this period of time is not less than the total amount of echo data that all detectors can respond to and output within this period of time.

Figure 9a shows the connection relationship between the four processing units 13 sharing the echo data of the even-numbered BANK. According to the technical concept of the present invention, the connection can also be carried out according to the embodiments of Figure 9b and Figure 9c, and the connection with the embodiment of Figure 9a The differences between the embodiments are: first, Figure 9b shares the echo data of even-numbered BANKs through three processing units, and Figure 9c shares the echo data of even-numbered BANKs through five processing units. If we only pursue to use as few processing units as possible to cope with As the point cloud data rate changes, Figure 9b is optimal compared to Figures 9a and 9c. Secondly, Figure 9c looks from the perspective of the processing unit. 13C and 13D in the middle are paired, and the two processing units 13 are connected in the same way. BANK 6, BANK 8 and BANK 10, which means that 13C and 13D help each other and share the echo data from these three BANKs. 13A and 13E on the upper and lower edges take care of themselves and are not paired with other processing units 13. Part of the BANK they are connected to is located in the second area. The amount of data is relatively small and the processing of echo data can be completed independently. Therefore, a paired processing unit 13 is set for a BANK that has a relatively large amount of echo data within the scan cycle (one scan cycle may include one or more transmissions and corresponding reception processes, which will take a certain amount of time), Thus, the effect of mutual assistance can be achieved. If we only focus on this point, Figure 10c is better than the solutions of Figure 9a and Figure 9b. However, in fact, the number of processing units 13 and the connection relationship between BANK and the processing unit 13 also need to consider other limiting factors, which will be further described below.

According to a preferred embodiment of the present invention, the processing unit 13 connected to each detection unit 12 is determined according to the echo data processing duration of each detection unit 12 .

The number of processing units 13 and the connection relationship between BANK and the processing unit 13 are related to the maximum load of each processing unit 13 or the number of echoes of the detection channels that can be processed. If the processing time of the processing unit 13 is divided into time slices, the time slice corresponds to the echo processing time given to each detection channel, specifically including detection channel switching time, encoding time, and flight time. In long-range radar, flight time accounts for the largest proportion. For example, in a radar with a ranging range of 300m, it is necessary to allocate hundreds of ns (channel switching time) + hundreds of ns (encoding time) + 2000ns (flight time = 300m×2m/3×10 ⁸ m/s) = up A time slice of thousands of ns is given to each detection channel. Therefore, when determining the connection relationship between the detection unit 12 and the processing unit 13, the duration of echo data processing needs to be considered.

According to a preferred embodiment of the present invention, when determining the processing unit 13 connected to each detection unit 12, the rotation speed and point cloud resolution of the lidar 10 also need to be considered.

The number of processing units 13 and the connection relationship between BANK and the processing unit 13 are also limited by the rotation speed and point cloud resolution of the radar. Once the speed and resolution of the radar are fixed, for example, the horizontal resolution is 0.1°, the vertical resolution is 0.2°, corresponding to x detection channels required for detection, it is required that all e detection channels must be sent and received within the time the radar rotates 0.1°. The faster the rotation speed, the less time is given to each detection channel, and f (f<e) detection channels may be detected in parallel. Multiple processing units 13 must share with each other to complete the echo data of these f channels. processing. For example, the radar scans the horizontal field of view driven by the rotating mirror, and the scanning frequency is 5Hz. If it needs to rotate g° (relatively rough horizontal angular resolution), it requires (1s/5×360°)×g° = h us, then h us is a scan cycle. If you need to rotate i° (fineer horizontal angle resolution, i<g), j us is required, and j us is a scanning period.

According to the planning of ranging and timing, the time slice required for each detection channel to complete the detection is d us (= the length of time given to complete the processing of the echo data of one detection channel). Therefore, within the scanning period of h us, each The maximum point cloud load of processing units 13 is h/d.

If the user has high requirements for the resolution of the tracking area, for example, within the h us, the number of point clouds that need to be detected in each area is determined. For the first area, the wandering range is n lines, and the BANK of the first area is Corresponding to the mode index, it is calculated that a fully loaded BANK in the first area needs to complete h/d point clouds in hus, which is the sum of the numbers distributed to all processing units 13 for processing. At the same time, this number is exactly equal to the maximum point cloud volume that each processing unit 13 can receive within hus. Therefore, for the BANK corresponding to the first area, one BANK may require one processing unit 13, and this processing unit 13 is unable to serve the processing of echo data of other BANKs.

Continuing to refer to Figures 9a-9c, it can be understood that the reason why banks 6, 8, and 10 in the center of the first area roaming range are mapped to multiple processing units 13 instead of one processing unit 13 is because the When an area is moving and reaches the peak data (for example, 36), the echo data of one BANK needs to occupy a processing unit 13. Of course, the echo data of another BANK needs to be divided into another or multiple processes. Unit 13 to process.

According to the above analysis, the situation of processing resources is: the processing unit 13 does not have a one-to-one correspondence with the BANK, and the number of processing units 13 is limited. The wiring rules in Figures 9a-9c are: each processing unit 13 is connected to multiple BANKs, and each BANK is connected The ultimate purpose of multiple processing units 13 is to ensure that no matter which area division or scanning mode (representing the distribution of echo and point cloud data volume), the detection of the corresponding work is The echo data output by the unit 12 can be processed by the relatively limited processing unit 13 .

Theoretically, each BANK can be connected to a processing unit, and each processing unit can be connected to each BANK. However, preferably, one factor can also be considered, namely: layout layout.

According to a preferred embodiment of the present invention, the processing unit 13 connected to each detection unit 12 is determined according to the physical distance between the detection unit 12 and the processing unit 13 .

The number of processing units 13 and the connection relationship between BANK and the processing unit 13 are also limited by the layout. Specifically, when multiple BANKs and multiple processing units 13 are laid out on a circuit board, the more lines that need to be connected, the more complicated it becomes. If operated according to theory, each detection unit 12 is connected to all processing units, and Each processing unit 13 is connected to the detection unit, but it is limited by: 1) The circuit board area is limited and not enough to arrange all the connections; 2) The connections may cross each other, causing signal crosstalk; 3) Heat dissipation is not smooth , poor reliability.

Therefore, the manifestation of this limiting factor is that each BANK or processing unit should be connected to the processing unit or BANK within a relatively close physical distance to itself, and try to avoid too many connections or too many cross lines.

The above introduces the limiting factors of the number of processing units and the connection relationship between BANK and processing units. In the introduction to area division and scanning mode, it is mentioned that the activable area has 9 scanning modes. In each scanning mode, The coverage range of the activateable area is different. Correspondingly, the internal control timing and point cloud format are different, and the activation methods between the BANK and the processing unit can be different. Therefore, after comprehensively considering the above constraints and completing the connection between the BANK and the processing unit, we can then plan how to combine each detection within this period based on the amount of data in each scanning mode and the amount of data that the MUX can bear. The echo data output by the processor is allocated to an appropriate amount of processing resources to complete point cloud calculation and generation. This allocation must ensure that the amount of echo data allocated to each processing unit is not larger than the amount of echo data that the processing unit can process within the time period. The upper limit of the number of echoes that can be processed must also ensure that the total amount of echo data processed by all processing units within this period is not less than the total amount of echo data output by all detectors.

Figure 10a shows a schematic diagram of the connection and echo data distribution method of the scanning mode index=1 according to an embodiment of the present invention. In the figure, the diagonal filled line represents the current first area, and the dotted line represents the hardware connection. Relationship, the solid line represents the actual connection to which echo data is allocated, and the number on each solid line represents the number of point clouds completed by the corresponding processing unit in one cycle, such as h=90us. Figure 10a only lists the design of even-numbered banks. The odd-numbered banks are symmetrical designs and will not be described or shown here. The number of point clouds in the first area in both the horizontal field of view and the vertical field of view is g=3 times that of other areas in the first area's wandering range except the current first area, so the total number is g ² =9 times. Since the point cloud volume of BANK 4 and BANK 6 reaches the peak (for example, 36), each must correspond to a processing unit 13, that is, 13B processes all the echo data of BANK 4, and 13C processes all the echo data of BANK 6. While other BANKs have smaller point cloud volumes, 13A, 13D and 13E can each correspond to two BANKs, which means that the total amount of echo data from the detectors of the two BANKs can be processed within this period of time.

Figure 10b shows a schematic diagram of the connection and echo data distribution method of the scanning mode index=9 according to another embodiment of the present invention. In the figure, the diagonal filled line represents the current first area, the dotted line represents the hardware connection relationship, and the solid line Indicates the actual connection to which echo data is allocated. The number on each solid line indicates the number of point clouds completed by the corresponding processing unit in one cycle. Taking the same even-numbered BANK as an example, since the point cloud volume of BANK 10 and BANK 12 reaches the peak (for example, 36), they must each correspond to a processing unit 13. The amount of point clouds in other BANKs is small, and 13A, 13B, and 13E can share the processing, that is, the total amount of echo data from the detectors of the three BANKs 13A, 13B, and 13E can be processed within this period of time.

According to a preferred embodiment of the present invention, multiple mapping tables are preset according to point cloud distribution. The mapping table includes the range of the first area and the echoes output by all detection units within this period of time based on the amount of echo data. The total amount of data and the upper limit of echoes (processing resources) that each detection unit can process within this period of time are used to allocate corresponding processing resources to each detection unit 12 in the first area.

Different cross-MUX mapping tables are designed according to point cloud distribution, the processing time of each processing unit 13 is divided according to time slices, and the time slices are allocated according to the MUX mapping table, thereby realizing multiple processing units 13 as a resource pool. Specifically, time slices are allocated to different BANKs so that detection channels in different BANKs can have different point cloud data rates to support movement of the first area in the vertical direction (or in any direction). Among them, the detection channel occupied by the first area may generate at least twice as much point cloud data as the second area, and relatively more time slices are allocated to the BANK corresponding to the first area. Allocate relatively small time slices to BANK corresponding to other areas.

The mapping table includes the range of the first area and the amount of echo data, the total amount of echo data output by all detection units within the duration, and the upper limit of the echo that each detection unit can process within the duration (processing resource). For example, according to the preset scanning mode of the first area in the vertical direction, a fixed cross MUX mapping table is correspondingly set. The table contains multiple time slices, each time slice specifies the address of each cross-MUX, and the BANK internal tree MUX address, so that the detection channel of the work in each time slice can be determined, and then each point cloud can be determined working time, and the corresponding horizontal angle.

According to a preferred embodiment of the present invention, the matching mapping table is switched according to changes in point cloud distribution.

Different cross-MUX mapping tables are preset according to the point cloud distribution, and then according to changes in the point cloud distribution, a more matching mapping table can be selected and switched to achieve dynamic tracking of obstacles or target areas.

For example, the user specifies the index number of the scanning mode, and the lidar automatically switches the corresponding cross-MUX mapping table and point cloud format, so that the first area can be moved in the vertical direction in real time.

According to a preferred embodiment of the present invention, the mapping table is switched between two probes.

To sum up, the above embodiments have introduced the area division, scanning mode, point cloud distribution of each area, the hardware connection relationship between the processing unit 13 and the detection unit 12 and how to allocate processing resources. The present invention comprehensively considers Hardware overhead and algorithm implementation use as few measurement resources as possible to process the changes in resolution and point cloud data rate caused by dynamic tracking of one or more areas. It is especially suitable for engineering requirements for a balance between cost, volume and efficiency.

According to a preferred embodiment of the present invention, referring to Figure 3a, the lidar 10 includes a plurality of transmitting units 11, a plurality of detection units 12 and a plurality of processing units 13, the plurality of transmitting units 11 are configured to emit detection beams, The plurality of detection units 12 and the plurality of transmitting units 11 form a plurality of detection channels, and each detection channel corresponds to a vertical orientation. The present invention also provides a resource allocation method for laser radar. The resource allocation method includes:

The plurality of detection units 12 receive echoes of the detection beams sent by the transmitting unit 11 that are reflected by obstacles;

Distribute the echo data of the detection unit 12 to one or more processing units 13 connected thereto to process the echo and generate a point cloud;

The echo processing relationship between the detection unit and the processing unit can be dynamically adjusted.

According to a preferred embodiment of the present invention, at least one detection unit 12 is connected to multiple processing units 13 , and for a detection unit 12 connected to multiple processing units 13 , its echo data can be shared and processed by multiple processing units 13 .

According to a preferred embodiment of the present invention, the resource allocation method further includes: when the field of view of the lidar 10 includes a first area and a second area, and the amount of echo data in the first area is greater than When the amount of echo data in the second area is described, according to the amount of echo data, the detection unit 12 corresponding to the first area is allocated processing resources that can complete the total amount of echo data that the detection unit 12 can generate within this period of time.

According to a preferred embodiment of the present invention, the first area may include multiple sub-areas, and the resource allocation method further includes: according to the amount of echo data, allocate to the detection unit 12 corresponding to the first area the ability to complete the Processing resources for the total amount of echo data that the detection unit 12 can generate within a period of time.

According to a preferred embodiment of the present invention, the plurality of sub-regions do not overlap with each other.

According to a preferred embodiment of the present invention, the resource allocation method further includes: according to the amount of echo data, allocate to the detection unit 12 corresponding to the second area the echo that the detection unit 12 can generate within the period of time. Processing resources for the total amount of data.

According to a preferred embodiment of the present invention, the plurality of detection channels are divided into multiple groups, and the detection channels of each group emit light in a time-sharing manner. The resource allocation method further includes: according to the amount of echo data, the The detection unit 12 allocates processing resources that can complete the total amount of echo data that the detection unit 12 can generate within this period of time.

Multiple mapping tables are preset according to the point cloud distribution. The mapping tables include the range of the first area and the total amount of echo data output by all detection units within the period, and the ability of each detection unit to The upper limit of the echoes that can be processed within this period of time (processing resources) is used to allocate corresponding processing resources to each detection unit 12 in the first area. According to a preferred embodiment of the present invention, the resource allocation method further includes: switching matching mapping tables according to changes in point cloud distribution.

According to a preferred embodiment of the present invention, the resource allocation method further includes: switching the mapping table between two detections.

The present invention also relates to a computer-readable storage medium, including computer-executable instructions stored thereon, which when executed by a processor implement the resource allocation method as described above.

Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, it is still The technical solutions described in the foregoing embodiments may be modified, or some of the technical features may be equivalently replaced. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

A lidar including:

a plurality of transmitting units configured to transmit detection beams;

A plurality of detection units configured to receive echoes of detection beams emitted by the transmitting unit reflected by obstacles; and

A plurality of processing units, connected to the detection unit, configured to process the echoes received by the detection unit and generate point clouds;

Among them, the echo processing relationship between the detection unit and the processing unit can be dynamically adjusted.
The lidar of claim 1, wherein at least one of the detection units is connected to multiple processing units, and for a detection unit connected to multiple processing units, its echoes can be shared and processed by multiple processing units.
The lidar according to claim 1, when the field of view of the lidar includes a first area and a second area, and the amount of echo data in the first area is greater than the amount of echo data in the second area. At least one detection unit corresponding to the first area is connected to a plurality of processing units.
The lidar of claim 3, wherein the first area may include a plurality of sub-areas, and at least one detection unit corresponding to each sub-area is connected to a plurality of processing units.
The lidar of claim 4, wherein the plurality of sub-regions do not overlap with each other.
The lidar according to claim 3, wherein each detection unit corresponding to the first area is connected to a plurality of processing units.
According to the lidar of claim 3, each detection unit corresponding to the second area is connected to a processing unit.
The lidar of claim 1, wherein each processing unit is connected to a plurality of detection units.
The lidar according to any one of claims 1 to 8, wherein the lidar further includes a plurality of multiplexing units, the plurality of detection channels are divided into a plurality of groups, and the detection units of each group pass A multiplexing unit is connected to a processing unit, and multiple groups of detection units are connected to a processing unit through a multiplexing unit.
According to the lidar according to any one of claims 1 to 8, multiple scanning modes are preset corresponding to the expected point cloud results, and the minimum number of processing units that need to be connected to the detection unit to satisfy each scanning mode is determined.
According to the lidar of claim 10, the processing unit connected to each detection unit is determined according to the echo data processing time of each detection unit.
According to the lidar of claim 10, the processing unit connected to each detection unit is determined according to the rotation speed and point cloud resolution of the lidar.
According to the lidar of claim 10, the processing unit connected to each detection unit is determined according to the physical distance between the detection unit and the processing unit.
A resource allocation method for lidar. The lidar includes a plurality of transmitting units, a plurality of detection units and a plurality of processing units. The plurality of transmitting units are configured to emit detection beams. The resource allocation method includes:

The plurality of detection units receive echoes of the detection beams sent by the transmitting unit reflected by obstacles;

assigning the echo to one or more processing units connected to the detection unit to process the echo and generate a point cloud;

Among them, the echo processing relationship between the detection unit and the processing unit can be dynamically adjusted.
The resource allocation method according to claim 14, wherein at least one detection unit is connected to a plurality of processing units The processing unit is connected, and for a detection unit connected to multiple processing units, the echoes can be shared and processed by multiple processing units.
The resource allocation method according to claim 14, further comprising: when the field of view of the lidar includes a first area and a second area, and the amount of echo data in the first area is greater than the amount of echo data in the second area. When the amount of echo data is determined, the echoes in the first area are allocated to one or more processing units connected to the detection unit corresponding to the first area according to the amount of echo data.
The resource allocation method according to claim 16, wherein the first area may include a plurality of sub-areas, and the resource allocation method further includes: according to the amount of echo data in the multiple sub-areas, The echo data is distributed to one or more processing units connected to the detection units corresponding to the plurality of sub-areas.
The resource allocation method according to claim 17, wherein the plurality of sub-regions do not overlap with each other.
The resource allocation method according to claim 16, further comprising: allocating the echo data in the second area to the detection units connected to the corresponding second area according to the amount of echo data in the second area. processing unit.
The resource allocation method according to claim 14, wherein the plurality of detection channels are divided into multiple groups, and the resource allocation method further includes: dividing the echo data of each group of detection channels according to the amount of echo data of the detection channel. The data is distributed to one or more processing units connected to the set of detection channels.
The resource allocation method according to claim 16, further comprising: presetting a plurality of mapping tables according to point cloud distribution, the mapping table including the range of the first area and the first area according to the amount of echo data in the first area. The echo data in a region is distributed to the processing unit connected to the detection unit corresponding to the first region.
The resource allocation method according to claim 21, further comprising: according to changes in point cloud distribution, Switch the matching mapping table.
The resource allocation method according to claim 22, further comprising: switching the mapping table between two detections.
A computer-readable storage medium includes computer-executable instructions stored thereon, which when executed by a processor implement the resource allocation method according to any one of claims 14-23.