Distributed acceleration simulation device for radar signal processing prototype

Abstract

According to the radar signal processing prototype distributed acceleration simulation method, the real-time simulation requirement of a radar signal processing prototype under mass echo data streams is met, the defects in the prior art are overcome, a radar signal processing prototype simulation acceleration agent instance resource pool is pre-constructed through multi-core and multi-calculation-node calculation resources, and the simulation acceleration agent instance resource pool is used for simulating the radar signal processing prototype. Radar echoes collected in real time are preferentially scheduled to simulation acceleration agents with low loads, the simulation acceleration agents start radar signal processing prototypes to carry out echo data processing, and the simulation acceleration agents can run in parallel and process different echo data frames at the same time. And meanwhile, low-delay data distribution is realized by adopting multi-thread asynchronous communication and a shared memory, so that simulation acceleration of a radar signal processing prototype in a distributed computing environment is realized, on-demand elastic expansion and contraction are supported, and the simulation rate of the prototype meets the real-time processing performance requirement.

Description

Technical Field

[0001] The invention belongs to the technical field of signal processing, and in particular relates to a radar signal processing prototype distributed acceleration simulation device. Background Art

[0002] Radar signal processing prototypes are used to verify the functions and performance of signal processing algorithms during the radar design phase, and are usually implemented based on the Matlab scientific computing platform. Radar signal processing prototype simulations need to process the massive echo data streams generated by the simulator, which has high requirements for real-time performance. The traditional offline simulation method of radar signal processing prototypes based on serial execution logic has low computing resource utilization and long simulation cycles, and cannot meet the requirements of online analysis and real-time control of radar signal processing.

[0003] The parallel computing toolbox provided by the existing Matlab platform has limited support for parallel programming. Among them, the parfor and SPMD parallel primitives are used for offline data parallel processing, which requires intrusive modification of the program. Matlab distributed computing service supports running multiple Matlab instances on multiple computing nodes to execute tasks in parallel. It adopts a polling scheduling strategy, does not support real-time task scheduling under different resource configurations, and data distribution is based on a single-threaded synchronous model with high transmission delay. Therefore, the existing Matlab platform cannot meet the real-time processing requirements of echo data streams. Summary of the invention

[0004] The present invention is aimed at the real-time simulation demand of radar signal processing prototype under massive echo data stream, overcomes the shortcomings of the prior art, and proposes a distributed acceleration simulation device for radar signal processing prototype. The computing resources of multi-core and multi-computing nodes are used to pre-build a radar signal processing prototype simulation acceleration agent instance resource pool, and the real-time collected radar echoes are preferentially scheduled to the simulation acceleration agent with lower load. The simulation acceleration agent starts the radar signal processing prototype to process the echo data. Each simulation acceleration agent can run in parallel and process different echo data frames at the same time. At the same time, multi-threaded asynchronous communication and shared memory are used to realize low-latency data distribution, thereby realizing the simulation acceleration of the radar signal processing prototype in a distributed computing environment, supporting on-demand elastic scaling, and making the prototype simulation rate meet the real-time processing performance requirements.

[0005] The radar signal processing prototype distributed acceleration simulation device of the present invention comprises an echo collection and distribution scheduling module, a prototype simulation node management module, a prototype simulation acceleration agent module and a simulation result summary module.

[0006] The functions of the echo acquisition and distribution scheduling module are:

[0007] a) Receive real-time echo data sent by the simulator through the RocketIO protocol and store it in the cache.

[0008] b) Responsible for maintaining the computing node network, load and prototype simulation acceleration agent module information collected by the prototype simulation node management module.

[0009] c) Priority scheduling is performed according to the echo sequence number, and computing nodes with lower load and available prototype simulation acceleration agent module instances are selected.

[0010] d) Send the echo data point-to-point to the prototype simulation node management module through a multi-threaded asynchronous network model.

[0011] The functions of the prototype simulation node management module are:

[0012] a) Build and maintain a prototype simulation acceleration agent module resource pool for each computing node.

[0013] b) Collect and feedback computing node network, load and available prototype simulation acceleration agent module information.

[0014] c) Receive the echo data sent by the echo collection and distribution scheduling module and store it in the shared memory queue.

[0015] The functions of the prototype simulation acceleration agent module are:

[0016] Monitor the echo data in the shared memory queue, take out the echo data if any, start the radar signal processing prototype program to process the echo data, and send the results to the simulation result summary module.

[0017] The functions of the simulation result summary module are:

[0018] Receive the prototype processing results sent by the prototype simulation acceleration agent module, reorder the results according to the echo sequence number, and send them to the next level of processing in the echo order.

[0019] Furthermore, the prototype simulation acceleration agent module is implemented using Matlab, which can directly call the radar signal processing prototype and use Matlab C / C++ mixed programming to implement the operation of the shared memory queue.

[0020] Furthermore, the prototype simulation node management module registers with the echo collection and distribution scheduling module, starts the prototype simulation acceleration proxy module resource pool, and regularly feeds back the network, load and available prototype simulation acceleration proxy module resource information of the computing node to the echo collection and distribution scheduling module.

[0021] Furthermore, when the echo acquisition and distribution scheduling module receives the echo data, it stores the data in the FIFO cache queue, takes out the queue head data in turn in a loop and schedules it to the computing node with low load, and distributes it to the prototype simulation node management module through multi-threaded asynchronous transmission. After receiving the echo data, the prototype simulation node management stores it in the shared memory queue and notifies the prototype simulation acceleration agent module. The prototype simulation acceleration agent in the listening state takes out the data from the shared memory queue and calls the radar signal processing prototype for processing. The processing results are sent to the simulation result summary module, and the results are re-sorted and sent to the next level of processing in sequence.

[0022] Furthermore, when echo data processing loses frames, computing nodes are added as needed, the prototype simulation node management module of the newly added node starts the prototype simulation acceleration agent module resource pool, and registers with the echo collection and distribution scheduling module, thereby participating in the prototype simulation acceleration.

[0023] Furthermore, when the echo collection and distribution scheduling module detects that the instances of the prototype simulation acceleration agent module are relatively idle, the echo data are preferentially scheduled to a small number of computing nodes, and the idle computing nodes can exit.

[0024] The present invention makes full use of the computing resources of multiple cores and multiple computing nodes to realize distributed parallel processing of radar signal processing prototypes, supports elastic scaling on demand, and enables the simulation rate of the prototype to meet the requirements of real-time processing performance; the data exchange method based on asynchronous network transmission and shared memory has low transmission delay, opens up the link between real-time data and Matlab prototype processes, and enables the Matlab prototype to support streaming data processing; the streaming data is preferentially scheduled to computing nodes with lower network, computing and other loads, and supports Matlab applications with high real-time requirements; there is no need to modify the radar signal processing prototype program, and the method has universality and no restrictions on the programming language implemented.

[0025] The beneficial effects of the present invention are

[0026] 1. The present invention provides a distributed accelerated simulation method for a radar signal processing prototype, which fully utilizes the computing resources of multiple cores and multiple computing nodes to realize distributed parallel processing of the radar signal processing prototype, supports on-demand elastic scaling, and enables the prototype simulation rate to meet the real-time processing performance requirements.

[0027] 2. The data exchange method based on asynchronous network transmission and shared memory of the present invention has low transmission delay, opens up the link between real-time data and Matlab prototype process, and enables Matlab prototype to support streaming data processing.

[0028] 3. The method proposed in the present invention preferentially dispatches streaming data to computing nodes with lower network and computing loads, supporting Matlab applications with high real-time requirements.

[0029] 4. Compared with the existing Matlab program parallel computing method, the present invention does not need to modify the radar signal processing prototype program, has universal applicability, and has no restrictions on the implemented programming language. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a diagram of the distributed accelerated simulation framework of the radar information processing prototype of the present invention.

[0031] Figure 2 This is a processing flow chart of the echo collection and distribution scheduling module of the present invention.

[0032] Figure 3 This is a processing flow chart of the prototype simulation node management module of the present invention.

[0033] Figure 4 This is a processing flow chart of the prototype simulation acceleration agent module of the present invention.

[0034] Figure 5 It is a processing flow chart of the simulation result summary module of the present invention. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] like Figure 1-Figure 5 As shown, a radar signal processing prototype distributed acceleration simulation device provided by the present invention includes: an echo collection and distribution scheduling module, whose main functions are: a) receiving real-time echo data sent by the simulator through the RocketIO protocol and storing it in a cache; b) responsible for maintaining the computing node network, load and prototype simulation acceleration agent module information collected by the prototype simulation node management module; c) performing priority scheduling according to the echo sequence number, selecting computing nodes with lower load and available prototype simulation acceleration agent module instances; d) sending echo data point-to-point to the prototype simulation node management module through a multi-threaded asynchronous network model.

[0037] The prototype simulation node management module has the following main functions: a) building and maintaining a prototype simulation acceleration agent module resource pool for each computing node; b) collecting and feeding back information such as computing node network, load, and available prototype simulation acceleration agent modules; c) receiving echo data sent by the echo collection and distribution scheduling module and storing it in a shared memory queue.

[0038] The prototype simulation acceleration agent module monitors the echo data in the shared memory queue, takes out the echo data if there is any, starts the radar signal processing prototype program to process the echo data, and sends the result to the simulation result summary module. The prototype simulation acceleration agent is implemented using Matlab, which can directly call the radar signal processing prototype and use Matlab C / C++ mixed programming to operate the shared memory queue.

[0039] The simulation result summary module receives the prototype processing results from the prototype simulation acceleration agent module, reorders the results according to the echo sequence number, and sends them to the next level of processing in the echo sequence.

[0040] In order to further illustrate the specific implementation of the present invention, an example of a distributed accelerated simulation method for a radar signal processing prototype is given:

[0041] In the example, three high-performance Linux servers are prepared. The echo collection and distribution scheduling module and the simulation result summary module run on one server, and the other two servers run the prototype simulation node management module and the prototype simulation acceleration agent module.

[0042] The prototype simulation node management module of each computing node registers with the echo collection and distribution scheduling module, starts the prototype simulation acceleration agent module resource pool, and regularly feeds back the network, load and available prototype simulation acceleration agent resource information of the computing node to the echo collection and distribution scheduling.

[0043] When the echo acquisition and distribution scheduling module receives the echo data, it stores the data in the FIFO cache queue, takes out the queue head data in turn in a loop and schedules it to the computing node with low load, and distributes it to the prototype simulation node management module through multi-threaded asynchronous transmission. After receiving the echo data, the prototype simulation node management stores it in the shared memory queue and notifies the prototype simulation acceleration agent module. The prototype simulation acceleration agent in the listening state takes out the data from the shared memory queue and calls the radar signal processing prototype for processing. The processing results are sent to the simulation result summary module, and the results are re-sorted and sent to the next level of processing in sequence.

[0044] When echo data processing loses frames, computing nodes are added as needed. The prototype simulation node management module starts the prototype simulation acceleration agent module resource pool on the newly added node and registers with the echo acquisition and distribution scheduling module to participate in the prototype simulation acceleration.

[0045] When the echo collection and distribution scheduling module detects that the instances of the prototype simulation acceleration agent module are relatively idle, the echo data are preferentially scheduled to a small number of computing nodes, and the idle computing nodes can exit.

[0046] The present invention is not limited to the above specific implementations, and the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made to the above implementations based on the technical essence of the present invention shall be included in the protection scope of the present invention.

Claims

1. A radar signal processing prototype distributed acceleration simulation device, characterized in that: It includes echo acquisition and distribution scheduling module, prototype simulation node management module, prototype simulation acceleration agent module and simulation result summary module; The echo collection and distribution scheduling module: a) Receive real-time echo data sent by the simulator through the RocketIO protocol and store it in the cache; b) Responsible for maintaining the computing node network, load and prototype simulation acceleration agent module information collected by the prototype simulation node management module; c) Priority scheduling is performed according to the echo sequence number, and a computing node with a low load and an available prototype simulation acceleration agent module instance is selected; d) sending the echo data point-to-point to the prototype simulation node management module through a multi-threaded asynchronous network model; The prototype simulation node management module: a) Build and maintain a prototype simulation acceleration proxy module resource pool for each computing node; b) Collect and feedback computing node network, load and available prototype simulation acceleration agent module information; c) receiving echo data sent by the echo collection and distribution scheduling module and storing it in a shared memory queue; The prototype simulation acceleration agent module: Monitor the echo data in the shared memory queue, take out the echo data if any, start the radar signal processing prototype program to process the echo data, and send the result to the simulation result summary module; The simulation result summary module: Receive the prototype processing results sent by the prototype simulation acceleration agent module, reorder the results according to the echo sequence number, and send them to the next level of processing in the echo order.

2. The radar signal processing prototype distributed accelerated simulation method according to claim 1, characterized in that: The prototype simulation acceleration agent module is implemented using Matlab, which can directly call the radar signal processing prototype and use Matlab C / C++ mixed programming to operate the shared memory queue.

3. The radar signal processing prototype distributed accelerated simulation method according to claim 1, characterized in that: The prototype simulation node management module registers with the echo collection and distribution scheduling module, starts the prototype simulation acceleration proxy module resource pool, and regularly feeds back the network, load and available prototype simulation acceleration proxy module resource information of the computing node to the echo collection and distribution scheduling module.

4. The radar signal processing prototype distributed accelerated simulation method according to claim 1, characterized in that: When the echo acquisition and distribution scheduling module receives the echo data, it stores the data in the FIFO cache queue, takes out the queue head data in turn in a loop and schedules it to the computing node with low load, and distributes it to the prototype simulation node management module through multi-threaded asynchronous transmission. After receiving the echo data, the prototype simulation node management stores it in the shared memory queue and notifies the prototype simulation acceleration agent module. The prototype simulation acceleration agent in the listening state takes out the data from the shared memory queue and calls the radar signal processing prototype for processing. The processing results are sent to the simulation result summary module, and the results are re-sorted and sent to the next level of processing in sequence.

5. The radar signal processing prototype distributed accelerated simulation method according to claim 1, characterized in that: When echo data processing loses frames, computing nodes are added as needed, the prototype simulation node management module of the newly added node starts the prototype simulation acceleration agent module resource pool, and registers with the echo acquisition and distribution scheduling module to participate in the prototype simulation acceleration.

6. The radar signal processing prototype distributed accelerated simulation method according to claim 1, characterized in that: When the echo collection and distribution scheduling module detects that the instances of the prototype simulation acceleration agent module are relatively idle, the echo data are preferentially scheduled to a small number of computing nodes, and the idle computing nodes can exit.

Images