Today I changed the method of calling the hand detection agent. Originally, the hand detection agent is called by a thread that calls std::system() function. The problem here is that this method could not kill the agent when the main program is being terminated.

Now, I changed so that the agent is executed in an individual process using the fork() function. The fork() function creates a replica of the current process. New process gets the return value of 0 from the fork function, and old process gets the id of new process. Because I could essentially create new process and overwrite the program with execve() function, I can have more control of the process than the thread-based agent.

Original method :

void *hand_detection::ai_agent(void *data) {
    std::string str = "python3 hand_detection_python.py --debug 0 --video ";
    str += std::string(GlobalHandDetectionSystem::get_self()->video_device);
    std::system(str.c_str());
    // signal the main process that the program has ended
    ((bool*)data)[0] = false;
    pthread_exit(0x00);
}

New method :

/* (Removed the ai_agent() function) */

void hand_detection::execute_agent(void) {
    GlobalHandDetectionSystem *global_hand_info = GlobalHandDetectionSystem::get_self();
    std::cout << "waiting for the agent to send the data...\n";
    pid_t pid = fork();
    if(pid == 0) {
        execl("/usr/bin/python3" , "/usr/bin/python3" , "./hand_detection_python.py" , "--debug" , "0" , "--video" , global_hand_info->video_device , (char*)0x00);
        exit(0);
    }
    global_hand_info->agent_process_id = pid;
    
    ...

You can see that because the new agent is a new process, we can terminate the agent all as we want :

kill(global_hand_info->agent_process_id , SIGKILL);