JP5644060B2 - Data processing device - Google Patents

Description

  The present invention relates to a data processing apparatus that performs data processing by synchronizing a plurality of program modules.

  A system is known that performs complex data processing by combining a plurality of devices and program modules. When periodic data processing is performed in such a system, it is necessary to synchronize each device and program module and perform processing at an appropriate timing. In the technique described in Patent Document 1, time data is attached to output results from the first and second ECUs and transmitted to the third ECU, and the third ECU is connected to the first ECU and the first ECU. When the difference between the time data in the output results of the two ECUs is within a predetermined time, processing using them is performed.

JP 2005-2011144 A

In the technique of Patent Document 1, although it is possible to perform processing using data output within a predetermined time, for example, the processing of each ECU is not reliably synchronized. Although there is a method of setting these synchronization periods at the system design stage, it is necessary to review the setting of the synchronization period every time the system is updated or integrated, which is complicated. In addition, it is difficult to cope with the case where there are processes with different synchronization periods in the system or when there are processes with variable synchronization periods.

Therefore, an object of the present invention is to provide a data processing apparatus that can automatically set and control a synchronization period between program modules.

To solve the above problems, the data processing apparatus according to the present invention, among a plurality of program modules, the data process unit for exchanging data timestamp is assigned, in each program module, and exchanges the data based on the connection relationship with other program modules, provided with a single integrated program modules Ru is formed as a synchronization group program modules for performing a synchronization process by using the data between the same timestamp, the integration program The module is characterized in that the processing is performed by adjusting the processing timing of each program module so that each synchronization group operates in a synchronization cycle set based on the connection relationship .

Each program module individually has an identification code, and when the program module performs an activation process, each program module activates its own program module and other program modules that the program module wishes to connect to This formation may be performed by following the connection relationship by notifying the program module using the identification code .

  According to the present invention, the integration program module classifies the modules that require data synchronization based on the data transmission / reception connection relationship of the modules and integrates them together as a synchronization group. Therefore, even when processes having different synchronization cycles are included, each can be operated with an optimum synchronization cycle. Further, even when there is a process with a variable synchronization period, an operation corresponding to the change in the period is possible. Therefore, system update and integration are easy. By integrating the synchronization processing using the identification code, the integration processing becomes easy and convenience is improved.

It is a block diagram which shows the structure of 1st Embodiment of the data processor which concerns on this invention. It is a figure explaining a data structure. It is a figure explaining the subject of the synchronous process in the data processor which concerns on this invention. It is a figure explaining the synchronous group in the data processor concerning the present invention. It is a figure explaining the integration process by the program module for integration. It is a figure explaining the emergency stop notification function by the data processor concerning the present invention. It is a figure explaining the automatic stop propagation function by the data processor concerning the present invention. It is a figure explaining the operation period verification function by the data processor concerning the present invention. It is a figure explaining the hierarchy process of the maintenance function by the data processor which concerns on this invention.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of the invention will be described in detail with reference to the accompanying drawings. In order to facilitate the understanding of the description, the same reference numerals are given to the same components in the drawings as much as possible, and duplicate descriptions are omitted.

  FIG. 1 is a block diagram showing a configuration of a first embodiment of a data processing apparatus according to the present invention. This data processing apparatus is a data processing apparatus for vehicles, and performs pedestrian detection, posture estimation, etc. using feature point extraction based on an input image (moving image) from a camera (not shown). is there. Each of the program modules (PU: Processing Unit) 1 to 4, 7, and 9 is mainly configured by an algorithm execution unit that executes an arbitrary user algorithm. 13, 31, 32), and a data output unit that transmits data (11-13, 31, 32) in response to a request from another PU.

  Of these, PU7 has an integrated function that automatically connects other PUs (1 to 4, 9), traces the connection relationship between previous and next PUs based on the data exchange, and automatically generates a synchronization table. ing. PU1 receives an input image and performs preprocessing and the like, and its operation cycle is 30 ms. PU3 extracts feature points in the image, and its operation cycle is 50 ms. PU2 detects a pedestrian based on the feature point extraction result, and its operation cycle is 150 ms. The PU 4 estimates the posture of the host vehicle based on the feature point extraction result, and its operation cycle is 500 ms. The PU 9 displays arbitrary data of the image data processed by the PU 1.

  Based on the synchronization table automatically generated by the PU 7, in the above embodiment, the PUs 1 to 3 form one synchronization group 5 as Group 0, while the PUs 1, 3, and 4 form one synchronization group 6 as Group 1. Synchronization group 5 synchronizes with a period of 150 ms, while group 6 synchronizes with a period of 500 ms. The PU 9 operates independently of these in a separate cycle.

  Next, a specific integration method will be described. FIG. 2 is a diagram for explaining the data structure of each of the data 11 to 13, 31 and 32. Each data includes a header part H, a group ID description part G, and a data part D. In the header portion H, information for identifying the output source PU is described, and whether or not it is necessary to synchronize with other PUs is described. Hereinafter, a data processing apparatus as shown in FIG. 3 will be described as an example. This data processing apparatus is a system that performs automatic operation based on an image acquired by a camera 90. In this system, the PU 20 extracts feature points from the image, and using the feature point data, the obstacle detection by the PU 10 and the self-position estimation by the PU 30 are performed. The PU 40 determines the necessity of performing a collision avoidance operation with an obstacle, and executes a collision avoidance operation such as an emergency stop. The PU 50 creates a route plan for route guidance based on the current position and the set destination information.

  In this system, if the obstacle position determined by the obstacle determination and the current position estimated by the self-position estimation are not at the same time, correct determination cannot be performed in collision avoidance determination or route planning. Therefore, the PU 10 that performs obstacle detection and the PU 30 that performs self-position estimation need to be synchronized. On the other hand, the emergency stop process by the PU 40 can be executed in a shorter cycle. When this entire system is synchronized, it operates at 5 Hz, which is the self-position estimation period of the PU 30 with the slowest processing speed, but the synchronization speed is too slow for emergency avoidance.

  According to the present invention, as shown in FIG. 4, the PU 10 and PU 40 are synchronized with a synchronization group as shown in FIG. It synchronizes with a period of 5 Hz as another synchronization group. Specifically, a time stamp is assigned to data from the camera 90, and the same time stamp is inherited for data corresponding thereto. Synchronization can be achieved by using data of the same time stamp. For example, if the output data from the camera is image (1), image (2), image (3),..., Image (n), the feature point extraction processing by the PU 20 is half the operation cycle with respect to the camera 90. The input data is image (1), image (3), image (5),..., Image (2m−1), and the output data is feature point (1), feature point (3), feature point (5). , ..., the feature point (2m-1).

  Similarly, the input data of the obstacle detection process in the PU 10 includes an image (1), an image (3), an image (5),..., An image (2m + 1), a feature point (1), a feature point (3), and a feature. The point (5), ..., the feature point (2m + 1), and the output data are the obstacle (1), the obstacle (3), the obstacle (5), ..., the obstacle (2m + 1). On the other hand, since the operation period of self-position estimation in the PU 30 is 1/6 of that of the camera 90, the input data is the feature point (1), the feature point (7), the feature point (13), ..., the feature point (6j + 1). The output data is the current position (1), the current position (7), the current position (13), ..., the current position (6j + 1).

  Each process of PU40 and PU50 can be performed with an operation cycle faster than PU10 and PU30. In this case, input data with a desired time stamp may not be obtained. However, data corresponding to the input data with the desired time stamp may be obtained from the received input data by a predetermined interpolation process or prediction process and used. . Thereby, each PU can be synchronized efficiently.

  The integration process by the integration PU will be described with reference to FIGS. Here, each PU (15, 25, 35) is assigned an identification code. As shown in FIG. 5 (a), each PU (15, 25, 35), when it is first activated, is the PU that it wishes to activate and connect to the integration PU 55 (the PU requests a data request). Is notified using this identification code. When synchronization is not required, the identification code of another PU may not be notified, or a format may be used in which notification is given that synchronization is not required. The integration PU 55 traces the connection relationship of each PU based on the information of the PUs desired to connect from each PU in this way, and generates a synchronization table. The method of tracing the connection relationship of each PU may follow the direction of processing from the preprocessing side to the postprocessing side, or from the postprocessing side to the preprocessing side in the opposite direction to the processing direction. May be.

  Then, when the PU for which connection is desired has already been activated, the integration PU 55 notifies the PU (25, 35) to that effect using an identification code. Thereby, a network is formed between PUs. In the next and subsequent cycles, the integration PU 55 controls the activation timing of each PU based on the generated synchronization table to synchronize each PU.

  When any of the PUs can change the processing speed, the integration PU 55 can change the synchronization process in accordance with the change of the PU based on the synchronization table. For example, as a change in the processing speed, in the lane recognition process, the moving speed is high on a highway, so the white line is easily recognized to speed up the cycle, while on a general road, the cycle is speeded up. By lowering the time to avoid the recognition process, it is possible to recognize even a slightly stripped white line. Thus, even when the processing cycle of some PUs is changed, according to the present invention, each PU can be automatically synchronized and operated. Furthermore, according to the present invention, PUs that do not need to be synchronized with other PUs, such as PU 9 shown in FIG. 1, are operated independently, so that they can be operated efficiently.

  Hereinafter, some of the other functions in the data processing apparatus according to the present invention will be exemplified. FIG. 6 is a diagram for explaining an emergency stop notification function by the data processing apparatus according to the present invention. In this data processing apparatus, for example, the server 150 and the client 100 cooperate to perform processing, and the integration PU 151 operates on the server 150. The data processing PU is composed of a PU 152 operating on the server 150 and a PU 110 operating on the client 100. The diagnostic applications 153 and 113 are operating on the server 150 and the PU 100, respectively. The client diagnosis application 113 periodically communicates with the diagnosis application 153 of the server 150, but the diagnosis application 153 performs timeout monitoring.

  When the timeout does not occur, normal processing is continued. However, when the timeout occurs, the diagnostic applications 153 and 113 transmit signals to the operating PUs 152 and 110, respectively. In the example shown in FIG. 6, the signal S <b> 2 is transmitted from the diagnostic application 113 to the PU 110. In the PU 110 that has received the medical checkup, a process registered as an abnormal process is executed. In the example shown in FIG. 6, the main function 111 calls the abnormal process handler 112 to execute a predetermined abnormal process.

  In this way, by configuring the diagnostic application as a program different from the PU, even when the PU becomes unresponsive, the operation abnormality can be detected. In addition, by not causing each PU to have a diagnostic function but by running a diagnostic application for each client and server, useless communication can be prevented and communication overflow can also be prevented.

  FIG. 7 is a diagram for explaining the automatic stop propagation function by the data processing apparatus according to the present invention. When a plurality of modules are operating in conjunction with each other, for example, if the downstream PU-H 208 is stopped for some reason, the PU that sends data only to the series connected to this PU-H 208 (this book In the example, PU-D204, PU-F206, and PU-G207 are applicable). If these PUs continue to drive, network resources, CPU resources and the like are consumed, which is not preferable. Therefore, each PU checks the status of the transmission destination PU at the time of data output, and if all the PUs at the transmission destination are stopped or in an abnormal state, the PU also stops its operation. As a result, in the main bell, the PU-D 204 and PU-G 207 confirm the status 210 of the PU-H 208, confirm the stop of the PU-H 208, describe that in their own status 211, 212, and operate. Stop. The PU-F 206 confirms the status 212 of the PU-G 207, confirms the stop of the PU-G 207, describes that fact in its own status 213, and stops the operation. On the other hand, since the PU-C 203 and PU-E 205 are operating, the PU-B 202 does not stop operating. In this way, it is possible to prevent consumption of network resources and CPU resources by stopping unnecessary PU operations based on the connection relationship.

  FIG. 8 is a diagram for explaining the operation cycle verification function by the data processing apparatus according to the present invention. In this data processing device 300, PUs 302 to 304 are operating under the integration PU 301. The integration PU 301 has a PU health check function 306, and checks data such as processing delay of each PU, time stamp of output data, number of waiting PUs, etc. (referred to as PU status data). Specifically, the integration PU 301 analyzes the collected PU status data to monitor the operation cycle of each PU, and executes error processing such as stopping an arbitrary PU if the delay is large. To do. Note that the emergency stop function 305 may be provided independently of the integration PU 301 and the PUs 302 to 304. By evaluating the processing cycle and the soundness of the processing by the integration PU 301, it is easy to confirm whether the operation efficiency expected by the designer is obtained, and the design efficiency can be improved. In addition, it becomes easy to achieve the operation cycle as expected by the designer, and system failure can be effectively prevented.

  FIG. 9 is a diagram for explaining the hierarchical processing of the maintenance function by the data processing apparatus according to the present invention. The data processing apparatus according to the present invention can also be configured by a network. And when the system of a data processing apparatus becomes a large-scale thing, there exists a possibility that the communication of the health check function by PU for integration may press the network. Therefore, instead of directly collecting the PU status data of each of the PUs 311 to 317, aggregation modules (aggregators) 321 to 324 are provided to distribute network communication, thereby reducing communication addition such as a health check function. It can be applied to a scale system.

  The data processing apparatus according to the present invention is suitable not only for a data processing apparatus based on a program executed by a single computer and control apparatus, but also for a data processing apparatus that operates in combination with a plurality of computers and control apparatuses.

  1-4, 9, 10, 15, 20, 25, 30, 35, 40, 50, 110, 152, 201-208, 302-304, 311-317 ... PU (Program Module, Processing Unit), 5, 6 ... Synchronization group, 7, 55, 301 ... PU for integration, 11-13, 31, 32 ... Data, 90 ... Camera, 100 ... Client, 111 ... Main function, 112 ... Processing handler in case of abnormality, 113, 153 ... For diagnosis Application, 150 ... server, 210-213 ... status, 300 ... data processing device, 305 ... emergency stop function, 306 ... PU health check function, 321-324 ... aggregation module.

Claims (2)

In a data processing apparatus that exchanges data with a time stamp assigned between a plurality of program modules,
In each program module, based on the connection relationships with other program modules for exchanging said data, one of integration of Ru was formed as a synchronous group program modules for performing a synchronization process by using the data together with the same time stamp A program module,
The data processing apparatus, wherein the integration program module adjusts the processing timing of each program module so that each synchronization group operates in a synchronization cycle set based on the connection relationship . Each program module individually has an identification code, and the integration program module, when each program module performs a startup process , activates its own program module and connects to its own program module. The data processing apparatus according to claim 1, wherein the formation is performed by following the connection relationship by notifying the program module using the identification code .

Images