uORB消息机制

官网英文原文地址:https://dev.px4.io/advanced-uorb.html

简介

uORB是一种用于进程间进行异步发布和订阅的消息机制API。

教程中可以学习通过C++如何使用uORB。

由于很多应用都是基于uORB的,因此在系统刚启动时uORB就自动运行了。uORB通过uorb start启动。可以使用uorb test进行单位测试。

添加新的主题(topic)

要想增加新的topic,你需要在msg/目录下创建一个新的.msg 文件并在msg/CMakeLists.txt下添加该文件。这样C/C++编译器自动在程序中添相应的代码。

可以先看看现有的msg文件了解下都支持那些类型。一个消息也可以嵌套在其他消息当中。

每一个生成的C/C++结构体中,会多出一个uint64_t timestamp 字段。这个变量用于将消息记录到日志当中。

为了在代码中使用”topic”需要添加头文件:

  1. #include <uORB/topics/topic_name.h>

首先需要在文件.msg中,通过添加类似如下的一行代码,一个消息定义就可以用于多个独立的主题.

  1. # TOPICS mission offboard_mission onboard_mission

【按】这里这一步将产生三个主题ID- mission、 offboard_mission 、以及 onboard_mission (第一个ID务必与.msg文件名相同)

然后在代码中, 把它们作为主题ID用:ORB_ID(offboard_mission).

发布主题

在系统的任何地方都可以发布一个主题, 包括在中断上下文中(被hrt_call接口调用的函数). 但是, 公告(advertise)一个主题仅限于在中断上下文之外.

一个主题只能由同一个进程进行公告, 并作为其之后的发布(publish).

列出所有主题并进行监听

接收者(listener)命令仅在Pixracer(FMUv4)以及Linux/OS X上可用。

要列出所有主题, 先列出文件句柄:

  1. ls /obj

要列出一个主题中的5个消息, 执行以下监听命令:

  1. listener sensor_accel 5

得到的输出就是关于该主题的n次内容:

  1. TOPIC: sensor_accel #3
  2. timestamp: 84978861
  3. integral_dt: 4044
  4. error_count: 0
  5. x: -1
  6. y: 2
  7. z: 100
  8. x_integral: -0
  9. y_integral: 0
  10. z_integral: 0
  11. temperature: 46
  12. range_m_s2: 78
  13. scaling: 0
  14. TOPIC: sensor_accel #4
  15. timestamp: 85010833
  16. integral_dt: 3980
  17. error_count: 0
  18. x: -1
  19. y: 2
  20. z: 100
  21. x_integral: -0
  22. y_integral: 0
  23. z_integral: 0
  24. temperature: 46
  25. range_m_s2: 78
  26. scaling: 0

uorb top Command

The command uorb top shows the publishing frequency of each topic in
real-time:

  1. update: 1s, num topics: 77
  2. TOPIC NAME INST #SUB #MSG #LOST #QSIZE
  3. actuator_armed 0 6 4 0 1
  4. actuator_controls_0 0 7 242 1044 1
  5. battery_status 0 6 500 2694 1
  6. commander_state 0 1 98 89 1
  7. control_state 0 4 242 433 1
  8. ekf2_innovations 0 1 242 223 1
  9. ekf2_timestamps 0 1 242 23 1
  10. estimator_status 0 3 242 488 1
  11. mc_att_ctrl_status 0 0 242 0 1
  12. sensor_accel 0 1 242 0 1
  13. sensor_accel 1 1 249 43 1
  14. sensor_baro 0 1 42 0 1
  15. sensor_combined 0 6 242 636 1

The columns are: topic name, multi-instance index, number of subscribers,
publishing frequency in Hz, number of lost messages (all subscribers combined), and
queue size.

Multi-instance

uORB provides a mechanism to publish multiple independent instances of the same
topic through orb_advertise_multi. It will return an instance index to the
publisher. A subscriber will then have to choose to which instance to subscribe
to using orb_subscribe_multi (orb_subscribe subscribes to the first
instance).
Having multiple instances is useful for example if the system has several
sensors of the same type.

Make sure not to mix orb_advertise_multi and orb_advertise for the same
topic.

The full API is documented in
src/modules/uORB/uORBManager.hpp.

Troubleshooting and common Pitfalls

The following explains some common pitfalls and corner cases:

  • The topic is not published: make sure the ORB_ID()‘s of each call match. It
    is also important that orb_subscribe and orb_unsubscribe are called from
    the same task
    as orb_publish. This applies to px4_task_spawn_cmd(), but
    also when using work queues (work_queue()).
  • Make sure to clean up: use orb_unsubscribe and orb_unadvertise.
  • A successful orb_check() or px4_poll() call requires an orb_copy(),
    otherwise the next poll will return immediately.
  • It is perfectly ok to call orb_subscribe before anyone advertised the topic.
  • orb_check() and px4_poll() will only return true for publications that are
    done after orb_subscribe(). This is important for topics that are not
    published regularly. If a subscriber needs the previous data, it should just
    do an unconditional orb_copy() right after orb_subscribe() (Note that
    orb_copy() will fail if there is no advertiser yet).