Based on the existing high-concurrency solutions, this paper studies the high-concurrency solutions for intelligent connected vehicle race based on cloud computing, and verifies the effectiveness of the scheme by designing the scheduling method of the intelligent networked car race platform and high-concurrency tests. The experimental results show that the designed scheduling method of the intelligent networked car event platform can meet the high concurrency requirements of the event, especially when international communication is affected by the epidemic, ensuring the participation of domestic and foreign teams in the competition and effective technical exchanges. It has a relatively broad application prospect.
tion, in order to meet the fairness, fairness and openness of the competition, it is necessary to ensure that the competition environment of the participants is consistent.Therefore, it is necessary to invent a high concurrency solution of intelligent connected auto race based on cloud computing.This can ensure the success of the intelligent connected car event, promote the technical exchanges among the participating teams, and help the implementation of domestic intelligent connected technology.
Based on the existing high-concurrency solutions, this paper studies the high-concurrency
solutions of intelligent connected auto races based on cloud computing. By designing the scheduling
method of the intelligent connected vehicle event platform[
In this experimental design, teams need to enter the corresponding contest interface by using their account and password on the competition login interface. After confirming the information on the question interface, manually start the question. In this case, the contest server automatically queries the current K8s cluster resource usage and determines whether there are idle resources available. If it determines that there are no idle resources available, the server directly enters the queuing mechanism. If it is determined that there are idle resources available, the server directly starts the virtual engine (UE4 is used in this experiment) and determines whether UE4 is successfully started. If the startup fails, the monitoring page is displayed. If the startup is successful, the server accounts for a period of time and checks whether an algorithm is connected within this period. If no algorithm is added, the server automatically stops the UE4 running process. If an algorithm is connected, the server will directly run the questions until the end of the questions and calculate the results.(see Figure 1).
In this experiment, the queuing mechanism is applied to the case that the server used in the contest determines that there is no idle resource available after automatically querying the current K8s cluster resource usage.
When the server automatically checks and determines that no idle resources are available, the current team will directly enter the queue, and the number of people in the queue will be increased by one. The current queue number is displayed. After that, it is up to the teams to choose whether to wait. If the team chooses not to wait, it will exit the queue directly, and the number of people in the queue will be increased or decreased by one. If you choose to continue waiting, the interface updates the number of people in line in real time.
When the server automatically queries that there are idle resources in the current K8s cluster resources, the team can choose whether to start the contest. If a team fails to manually click the start button in time within the time specified by the system, the team will directly exit the queue, lose the qualification and need to queue up again. If the team manually clicks the start button in time within the time specified by the system, the server will directly start UE4, and the current queue of the team will end (Figure 2).
This experiment adopts the container arrangement scheme of Kubernetes(K8s)[
In addition, K8s could manage multiple management nodes [
In this experiment, the algorithm access scheme design is applicable to the server accounting time for a period of time after UE4 is successfully started. And judge whether there is an algorithm access in this period of time.
In this experiment, the algorithm invokes the login interface for authentication. At the same time, verify that the user name and password are valid, and verify that the user has a running question. If the preceding authentication fails, the interface returns the failure cause, and the algorithm connection process ends. If the above verification is successful, the algorithm will call the Web interface to obtain the access key and the communication address. Then, the Web interface is called to obtain the installation sensor information of the main vehicle. Next, the algorithm invokes the UE4 interface with the secret key to send the start instruction. Next, the algorithm calls UE4 interface to obtain sensor data and master vehicle data, makes decisions, and sends control instructions according to the decisions. Finally, after running the questions, the questions are finished and the scores are calculated(Figure 4) .
In this experiment, after completing the design of the scheduling method for the intelligent
connected car race platform, it is necessary to carry out a high concurrency test on this method, so as to
verify that this experiment can finally solve the high concurrency requirements of intelligent
connected car race based on cloud computing. Therefore, the high concurrency test is carried out from the
following two aspects[
This experiment first needs to test how many UE4 can run on a single server. According to the requirements of this experimental competition, different kinds of sensors are installed for different kinds of questions. For example, only millimeter wave radar is installed in the decision control class, and only camera is installed in the perception class. Therefore, in this experiment, the decision control and perception questions are respectively run on the same K8s server, and the running conditions of the two different questions are represented by the number of UE4 runs. (as shown in Table 1)
After determining the number of UE4s running on a single K8s server, a high concurrency test experiment is required based on the overall race high concurrency requirements.（The high concurrency test environment required by the experiment is shown in Table 2 and Table 3） For example, in this experiment, 100 UE4s are run in the same time period. Therefore, this experiment will set up a test environment and carry out high concurrency tests for 100 UE4s running in the same time period from four aspects: time characteristics, resource utilization, capacity and compliance of performance efficiency(Figure 5) .
Test content The corresponding time, processing time and throughput when the product performs its functions under specified conditions.
CPU utilization and memory utilization when executing test tasks.
Whether the maximum number of concurrent users, communication bandwidth, transaction throughput and other parameters meet the requirements.
Whether the product complies with the requirements specification, product description, etc., as well as the performance and efficiency requirements in the standard.
Based on the scheduling method of intelligent connected car race platform, this method is tested with high concurrency in this experiment. The individual server concurrency was first tested to determine the amount of UE4 a single server could run. This was followed by a high concurrency test for large-scale events (this experiment required running 100 UE4 high concurrency tests in the same time period). Finally, after several rounds of testing and code tuning, the experimental results are as follows:
According to the different question types of the competition, this experiment selected six types of scenes in the competition: stationary front vehicle (straight line), pedestrian crossing the road (no cover), curved road, obstructed front vehicle, horizontal parking, vertical parking, automatic driving (low dynamic traffic flow). After 8 tests, the test results are shown in Figure 6. Marked red indicates that the server runs at a frame rate below 20fps, and marked green indicates that the server runs at a frame rate above 20fps.
In order to further guarantee the smooth operation of a single K8s server, this experiment further averages the number of UE4 running in the above different scenarios. In addition, if the frame rate of a single server is above 24fps, it is qualified. Therefore, after calculation, the number of UE4 questions in decision control category and running perception category can be run by a single server is 6 and 5 respectively (as shown in Table 4).
According to the requirement that a single K8s server can run 5 UE4, 20 K8s servers are selected for the high concurrency test of 100 concurrency. Then the monitoring data is randomly selected and the maximum value is recorded. After several rounds of testing and code tuning, the monitoring data is extracted as follows:
This highly concurrent test involves a total of 5 interfaces (system login, running the contest, obtaining the data of the main vehicle, obtaining the monitoring data, and the algorithm). In addition, 100 concurrent tests can be started in both the running contest and the data acquisition history, so the concurrency test passes (as shown in Table 5).
In this paper, the high concurrency solution of intelligent connected auto race based on cloud computing is studied. By designing the scheduling method of the intelligent connected car race platform, and through high concurrency test, the validity and reliability of the scheme are verified. The competition is further ensured through the cloud server, which guarantees the high concurrent demand of the intelligent connected car event based on cloud computing for competitors from different regions to go online at the same time. At the same time, it also ensures that all teams can see the ranking of each team in real time. In addition, ensure that the competition is fair, impartial and open. Especially when international exchanges are affected by the epidemic, it guarantees the participation of domestic and foreign teams in competitions and effective technical exchanges, and has a broad application prospect.