ROS应用层通信协议解析

时间:2022-11-19 17:05:09

参考:http://wiki.ros.org/ROS/Master_API

http://wiki.ros.org/ROS/Connection Header

ROS本质上就是一个松耦合的通信框架,通信模式包括:远程调用(service-client)、订阅发布(topic)、持续通信(action)和全局参数(参数服务器),这四种模式基本已经能够涵盖百分之九十的应用场景了

本次针对订阅发布模式,探究一下ROS通信中的具体通信协议,读完本文后,你可以在不依赖ROS的情况下和ROS通信

本次通信采用从机订阅主机数据,通过wireshark抓包,得到具体xmlrpc协议数据内容,根据xmlrpc协议格式,找到对应代码

(因为时间有限,部分协议可能有跳过的地方)

1、registerPublisher

从机创建节点的第一步

这个方法用于注册一个发布者的caller

request报文body:

<?xml version="1.0"?>
<methodCall>
    <methodName>registerPublisher</methodName>
    <params>
        <param>
            <value>/test_sub</value>
        </param>
        <param>
            <value>/rosout</value>
        </param>
        <param>
            <value>rosgraph_msgs/Log</value>
        </param>
        <param>
            <value>http://192.168.1.150:40209</value>
        </param>
    </params>
</methodCall>

response报文body:

<?xml version='1.0'?>
<methodResponse>
    <params>
        <param>
            <value>
                <array>
                    <data>
                        <value>
                            <int>1</int>
                        </value>
                        <value>
                            <string>Registered [/test_sub] as publisher of [/rosout]</string>
                        </value>
                        <value>
                            <array>
                                <data>
                                    <value>
                                        <string>http://sherlock:35861/</string>
                                    </value>
                                </data>
                            </array>
                        </value>
                    </data>
                </array>
            </value>
        </param>
    </params>
</methodResponse>

先说结论:

ROS有一个日志相关的topic,名称是 /rosout,所有节点的日志信息都会通过这个 topic 发布出来

ROS还有一个日志相关的节点,名称是 /rosout,负责订阅 /rosout数据,然后使用名称为 /rosout_agg 的topic发布出来, /rosout_agg 的订阅者是rqt等调试工具

所以,结合上面的xml内容,我们可以大致推断,创建一个新的节点,那么这个节点必定会发布一个topic,就是/rosout,所以上面的XMLRPC协议内容,就是网rosmaster内注册一个publisher,用于发布/rosout

整体来说,就是调用接口

registerPublisher("/test_sub", "/rosout", "rosgraph_msgs/Log", "http://192.168.1.150:40209")

返回值是:

1
Registered [/test_sub] as publisher of [/rosout]
http://sherlock:35861/

1是固定数据

第二行是message

最后一个返回值表示订阅者的uri列表,这里因为只有一个订阅者,所有只有一个uri

再看代码:

函数声明如下:

registerPublisher(caller_id, topic, topic_type, caller_api)
	Register the caller as a publisher the topic.
	参数
		caller_id (str)
			ROS caller ID
		topic (str)
			Fully-qualified name of topic to register.
		topic_type (str)
			Datatype for topic. Must be a package-resource name, i.e. the .msg name.
		caller_api (str)
			API URI of publisher to register.
	返回值(int, str, [str])
		(code, statusMessage, subscriberApis)
		List of current subscribers of topic in the form of XMLRPC URIs.

找到 registerPublisher 接接口,位于ros_comm/rosmaster 包中,文件为:master_api.py(ROS主从机制利用python实现,拿掉python则无法实现主从)

@apivalidate([], ( is_topic('topic'), valid_type_name('topic_type'), is_api('caller_api')))
def registerPublisher(self, caller_id, topic, topic_type, caller_api):
	"""
    Register the caller as a publisher the topic.
    @param caller_id: ROS caller id
    @type  caller_id: str
    @param topic: Fully-qualified name of topic to register.
    @type  topic: str
	@param topic_type: Datatype for topic. Must be a
    package-resource name, i.e. the .msg name.
	@type  topic_type: str
    @param caller_api str: ROS caller XML-RPC API URI
    @type  caller_api: str
    @return: (code, statusMessage, subscriberApis).
    List of current subscribers of topic in the form of XMLRPC URIs.
    @rtype: (int, str, [str])
    """
    #NOTE: we need topic_type for getPublishedTopics.
	try:
        self.ps_lock.acquire()
        self.reg_manager.register_publisher(topic, caller_id, caller_api)
        # don't let '*' type squash valid typing
        if topic_type != rosgraph.names.ANYTYPE or not topic in self.topics_types:
            self.topics_types[topic] = topic_type
        pub_uris = self.publishers.get_apis(topic)
        sub_uris = self.subscribers.get_apis(topic)
        self._notify_topic_subscribers(topic, pub_uris, sub_uris)
        mloginfo("+PUB [%s] %s %s",topic, caller_id, caller_api)
        sub_uris = self.subscribers.get_apis(topic)            
    finally:
    	self.ps_lock.release()
	return 1, "Registered [%s] as publisher of [%s]"%(caller_id, topic), sub_uris

registerPublisher 接口的注释:Register the caller as a publisher the topic,将调用者注册为一个topic发布者

可以对应xmlrpc中对应参数,加上猜测:

caller_id:调用者,可以认为是节点,/test_sub,从及创建的节点

topic:发布的topic name,/rosout

topic_type:发布的topic数据类型,rosgraph_msgs/Log

caller_api:调用者发布数据的API接口,http://192.168.1.150:40209

总上,我们大概有几点猜测:

  1. 接口在rosmaster中,接口是registerPublisher,表示,这是注册节点的
  2. 告诉master节点,我创建了一个节点,节点名是/test_sub
  3. 告诉master,这个节点需要发布topic,topic名是/rosout,数据类型是rosgraph_msgs/Log

registerPublisher 接口中有三个地方需要注意:

  1. register_publisher接口调用
  2. _notify_topic_subscribers接口调用,告知当前所有的subscriber,有新的publisher,他们需要再次到新的publisher中去订阅数据
  3. return 内容,最后会拼接成xmlrpc的报文,response 回去,这也就顺便解释了第二条xmlrpc报文(response)

register_publisher

先看 register_publisher,代码在rosmaster中的registrations.py文件中

def register_publisher(self, topic, caller_id, caller_api):
    """
	Register topic publisher
    @return: None
    """
    self._register(self.publishers, topic, caller_id, caller_api)

_register 接口,这个节点做了三件事

  1. 调用内部接口保存节点信息
  2. 如果这个节点之前已经存在,就表明它是在更新,则发布数据的接口改变,且之前已经有订阅,则此时所有订阅该接口的所有subscriber解除订阅
  3. 调用register接口保存
def _register(self, r, key, caller_id, caller_api, service_api=None):
    # update node information
    node_ref, changed = self._register_node_api(caller_id, caller_api)
    node_ref.add(r.type, key)
    # update pub/sub/service indicies
    if changed:
        self.publishers.unregister_all(caller_id)
        self.subscribers.unregister_all(caller_id)
        self.services.unregister_all(caller_id)
        self.param_subscribers.unregister_all(caller_id)
	r.register(key, caller_id, caller_api, service_api)

_register_node_api 接口,我们可以看到,它主要做两件事

  1. 更新master中节点信息(节点名、节点发布数据的接口)
  2. 检查这个节点是不是已经存在,如果是,则告诉调用者
def _register_node_api(self, caller_id, caller_api):
    """
    @param caller_id: caller_id of provider
    @type  caller_id: str
    @param caller_api: caller_api of provider
    @type  caller_api: str
    @return: (registration_information, changed_registration). changed_registration is true if 
    caller_api is differet than the one registered with caller_id
    @rtype: (NodeRef, bool)
    """
    node_ref = self.nodes.get(caller_id, None)

    bumped_api = None
    if node_ref is not None:
        if node_ref.api == caller_api:
            return node_ref, False
        else:
            bumped_api = node_ref.api
            self.thread_pool.queue_task(bumped_api, shutdown_node_task,
                                        (bumped_api, caller_id, "new node registered with same name"))

    node_ref = NodeRef(caller_id, caller_api)
    self.nodes[caller_id] = node_ref
    return (node_ref, bumped_api != None)

_notify_topic_subscribers

_notify_topic_subscribers 代码,根据注释说明,接口的作用就是通知所有的subscriber,有新的publisher

def _notify_topic_subscribers(self, topic, pub_uris, sub_uris):
    """
    Notify subscribers with new publisher list
    @param topic: name of topic
    @type  topic: str
    @param pub_uris: list of URIs of publishers.
    @type  pub_uris: [str]
    """
    self._notify(self.subscribers, publisher_update_task, topic, pub_uris, sub_uris)

_notify 代码,将更新的通知任务(publisher_update_task)放进事件队列中,等待执行:

def _notify(self, registrations, task, key, value, node_apis):
    """
    Generic implementation of callback notification
    @param registrations: Registrations
    @type  registrations: L{Registrations}
    @param task: task to queue
    @type  task: fn
    @param key: registration key
    @type  key: str
    @param value: value to pass to task
    @type  value: Any
    """
    # cache thread_pool for thread safety
    thread_pool = self.thread_pool
    if not thread_pool:
        return
    
    try:            
        for node_api in node_apis:
            # use the api as a marker so that we limit one thread per subscriber
            thread_pool.queue_task(node_api, task, (node_api, key, value))
    except KeyError:
        _logger.warn('subscriber data stale (key [%s], listener [%s]): node API unknown'%(key, s))

publisher_update_task 代码,传入的三个参数分别是:新节点的接口、topic名称、订阅者的接口:

def publisher_update_task(api, topic, pub_uris):
    """
    Contact api.publisherUpdate with specified parameters
    @param api: XML-RPC URI of node to contact
    @type  api: str
    @param topic: Topic name to send to node
    @type  topic: str
    @param pub_uris: list of publisher APIs to send to node
    @type  pub_uris: [str]
    """
    msg = "publisherUpdate[%s] -> %s %s" % (topic, api, pub_uris)
    mloginfo(msg)
    start_sec = time.time()
    try:
        #TODO: check return value for errors so we can unsubscribe if stale
        ret = xmlrpcapi(api).publisherUpdate('/master', topic, pub_uris)
        msg_suffix = "result=%s" % ret
    except Exception as ex:
        msg_suffix = "exception=%s" % ex
        raise
    finally:
        delta_sec = time.time() - start_sec
        mloginfo("%s: sec=%0.2f, %s", msg, delta_sec, msg_suffix)

publisherUpdate 接口在 rospy 模块的 masterslave.py 文件中,猜测是使用XMLRPC协议,通知所有的订阅者节点,发布者更新了

@apivalidate(-1, (is_topic('topic'), is_publishers_list('publishers')))
def publisherUpdate(self, caller_id, topic, publishers):
    """
    Callback from master of current publisher list for specified topic.
    @param caller_id: ROS caller id
    @type  caller_id: str
    @param topic str: topic name
    @type  topic: str
    @param publishers: list of current publishers for topic in the form of XMLRPC URIs
    @type  publishers: [str]
    @return: [code, status, ignore]
    @rtype: [int, str, int]
    """
    if self.reg_man:
        for uri in publishers:
            self.reg_man.publisher_update(topic, publishers)
    return 1, "", 0

2、hasParam

request报文body

<?xml version='1.0'?>
<methodResponse>
    <params>
        <param>
            <value>
                <array>
                    <data>
                        <value>
                            <int>1</int>
                        </value>
                        <value>
                            <string>/use_sim_time</string>
                        </value>
                        <value>
                            <boolean>0</boolean>
                        </value>
                    </data>
                </array>
            </value>
        </param>
    </params>
</methodResponse>

response报文body

<?xml version="1.0"?>
<methodCall>
    <methodName>requestTopic</methodName>
    <params>
        <param>
            <value>/rosout</value>
        </param>
        <param>
            <value>/rosout</value>
        </param>
        <param>
            <value>
                <array>
                    <data>
                        <value>
                            <array>
                                <data>
                                    <value>TCPROS</value>
                                </data>
                            </array>
                        </value>
                    </data>
                </array>
            </value>
        </param>
    </params>
</methodCall>

先说结论:

调用 hasParam 接口,检查参数服务器是否有参数 /test_sub/use_sim_time

返回值是 [1, /use_sim_time, 0],表示没有

再看代码:

hasParam

有了上面的分析经验,我们可以很轻松地的出结论,这是在调用 hasParam 接口

我们可以很轻松地找到这个方法的代码,在rosmaster模块的master_api.py文件中

@apivalidate(False, (non_empty_str('key'),))
def hasParam(self, caller_id, key):
    """
    Check if parameter is stored on server. 
    @param caller_id str: ROS caller id
    @type  caller_id: str
    @param key: parameter to check
    @type  key: str
    @return: [code, statusMessage, hasParam]
    @rtype: [int, str, bool]
    """
    key = resolve_name(key, caller_id)
    if self.param_server.has_param(key):
        return 1, key, True
    else:
        return 1, key, False 

根据协议

  1. caller_id 传参是 /test_sub
  2. key 传参是 /use_sim_time

根据注释和代码,我们可以确认,这个就口就是在检查,参数服务器是否有参数 /test_sub/use_sim_time

resolve_name 接口接收两个参数,根据调用:

  1. name是/use_sim_time
  2. namespace_是/test_sub

所以,才能确认上面的全局参数 /test_sub/use_sim_time

hasParam 接口的返回值有三个

  1. code,整型,这里无论有没有,都返回1,可以忽略
  2. key,这里就是 /use_sim_time
  3. hasParam,表示是否有这个参数,True/False

根据 response 报文,这里应该返回 [1, /use_sim_time, 0],表示没有这个参数

use_sim_time

想要理解为什么要调用这个接口,就要理解 use_sim_time 参数的作用

use_sim_time是一个重要的参数,它默认值为false,可以配合Rosbag使用,是一个很重要的离线调试工具

我们都知道,ROS 中的时间有两种:

  1. ROS::Time()
  2. ROS::WallTime()

ROS::Time()和ROS::WallTime()

表示ROS网络中的时间,如果当时在非仿真环境里运行,那它就是当前的时间。但是假设去回放当时的情况,那就需要把当时的时间录下来

以控制为例,很多的数据处理需要知道当时某一个时刻发生了什么。Wall Time可以理解为墙上时间,墙上挂着的时间没有人改变的了,永远在往前走;ROS Time可以被人为修改,你可以暂停它,可以加速,可以减速,但是Wall Time不可以。

在开启一个Node之前,当把use_sim_time设置为true时,这个节点会从clock Topic获得时间。所以操作这个clock的发布者,可以实现一个让Node中得到ROS Time暂停、加速、减速的效果。同时下面这些方面都是跟Node透明的,所以非常适合离线的调试方式。当把ROSbag记下来以后重新play出来时,加两个横杠,--clock,它就会发布出这个消息

3、registerSubscriber

先看报文:

request报文body

<?xml version="1.0"?>
<methodCall>
    <methodName>registerSubscriber</methodName>
    <params>
        <param>
            <value>/test_sub</value>
        </param>
        <param>
            <value>/ros_message</value>
        </param>
        <param>
            <value>my_package/MessageDefine</value>
        </param>
        <param>
            <value>http://192.168.1.150:43597</value>
        </param>
    </params>
</methodCall>

response报文body

<?xml version='1.0'?>
<methodResponse>
    <params>
        <param>
            <value>
                <array>
                    <data>
                        <value>
                            <int>1</int>
                        </value>
                        <value>
                            <string>Subscribed to [/ros_message]</string>
                        </value>
                        <value>
                            <array>
                                <data>
                                    <value>
                                        <string>http://sherlock:41689/</string>
                                    </value>
                                </data>
                            </array>
                        </value>
                    </data>
                </array>
            </value>
        </param>
    </params>
</methodResponse>

registerSubscriber

registerSubscriber 代码在rosmaster包中的master_api.py文件中,如下:

@apivalidate([], ( is_topic('topic'), valid_type_name('topic_type'), is_api('caller_api')))
def registerSubscriber(self, caller_id, topic, topic_type, caller_api):
    """
    Subscribe the caller to the specified topic. In addition to receiving
    a list of current publishers, the subscriber will also receive notifications
    of new publishers via the publisherUpdate API.        
    @param caller_id: ROS caller id
    @type  caller_id: str
    @param topic str: Fully-qualified name of topic to subscribe to. 
    @param topic_type: Datatype for topic. Must be a package-resource name, i.e. the .msg name.
    @type  topic_type: str
    @param caller_api: XML-RPC URI of caller node for new publisher notifications
    @type  caller_api: str
    @return: (code, message, publishers). Publishers is a list of XMLRPC API URIs
       for nodes currently publishing the specified topic.
    @rtype: (int, str, [str])
    """
    #NOTE: subscribers do not get to set topic type
    try:
        self.ps_lock.acquire()
        self.reg_manager.register_subscriber(topic, caller_id, caller_api)

        # ROS 1.1: subscriber can now set type if it is not already set
        #  - don't let '*' type squash valid typing
        if not topic in self.topics_types and topic_type != rosgraph.names.ANYTYPE:
            self.topics_types[topic] = topic_type

        mloginfo("+SUB [%s] %s %s",topic, caller_id, caller_api)
        pub_uris = self.publishers.get_apis(topic)
    finally:
        self.ps_lock.release()
    return 1, "Subscribed to [%s]"%topic, pub_uris

根据协议往来,我们可以看到调用过程

registerSubscriber("/test_sub", "/ros_message", "my_package/MessageDefine", "http://192.168.1.150:43597")

入参有4个:

  1. 订阅节点名:/test_sub
  2. 需要订阅的topic 名称:/ros_message
  3. topic的数据类型:my_package/MessageDefine
  4. 订阅节点自己的uri,即发布者通知时的发送目标

返回值有3个:

  1. code,这里固定是1

  2. message,这里是 Subscribed to [/ros_message]

  3. publisher 的订阅URI 列表,因为这里只有一个publisher,所以只有一个 http://sherlock:41689/,首先,这可能是主机里面某个几点的uri,需要从机去订阅

代码说明:

和第一条,注册publisher相反,这里是注册subscriber

有几个关键代码

register_subscriber 代码,位于rosmaster包中的registerations.py文件中:

def register_subscriber(self, topic, caller_id, caller_api):
    """
    Register topic subscriber
    @return: None
    """
    self._register(self.subscribers, topic, caller_id, caller_api)

_register 代码,调用 _register_node_api 接口更新节点信息,如果之前有该节点的注册信息,则先删除:

def _register(self, r, key, caller_id, caller_api, service_api=None):
    # update node information
    node_ref, changed = self._register_node_api(caller_id, caller_api)
    node_ref.add(r.type, key)
    # update pub/sub/service indicies
    if changed:
        self.publishers.unregister_all(caller_id)
        self.subscribers.unregister_all(caller_id)
        self.services.unregister_all(caller_id)
        self.param_subscribers.unregister_all(caller_id)
    r.register(key, caller_id, caller_api, service_api)

_register_node_api 代码:

def _register_node_api(self, caller_id, caller_api):
    """
    @param caller_id: caller_id of provider
    @type  caller_id: str
    @param caller_api: caller_api of provider
    @type  caller_api: str
    @return: (registration_information, changed_registration). changed_registration is true if 
    caller_api is differet than the one registered with caller_id
    @rtype: (NodeRef, bool)
    """
    node_ref = self.nodes.get(caller_id, None)

    bumped_api = None
    if node_ref is not None:
        if node_ref.api == caller_api:
            return node_ref, False
        else:
            bumped_api = node_ref.api
            self.thread_pool.queue_task(bumped_api, shutdown_node_task,
                                        (bumped_api, caller_id, "new node registered with same name"))

    node_ref = NodeRef(caller_id, caller_api)
    self.nodes[caller_id] = node_ref
    return (node_ref, bumped_api != None)

shutdown_node_task 代码,如果订阅节点退出了,则需要通知:

def shutdown_node_task(api, caller_id, reason):
    """
    Method to shutdown another ROS node. Generally invoked within a
    separate thread as this is used to cleanup hung nodes.
    
    @param api: XML-RPC API of node to shutdown
    @type  api: str
    @param caller_id: name of node being shutdown
    @type  caller_id: str
    @param reason: human-readable reason why node is being shutdown
    @type  reason: str
    """
    try:
        xmlrpcapi(api).shutdown('/master', "[{}] Reason: {}".format(caller_id, reason))
    except:
        pass #expected in many common cases
    remove_server_proxy(api)

4、requestTopic

request报文body,如下,我们可以看到,xmlrpc发送的host是sherlock:41689,是上一步骤,收到的publisher的uri,如果有多个publisher,要request多次

POST /RPC2 HTTP/1.1
Host: sherlock:41689
User-Agent: Go-http-client/1.1
Content-Length: 307
Content-Type: text/xml
Accept-Encoding: gzip

<?xml version="1.0"?>
<methodCall>
    <methodName>requestTopic</methodName>
    <params>
        <param>
            <value>/test_sub</value>
        </param>
        <param>
            <value>/ros_message</value>
        </param>
        <param>
            <value>
                <array>
                    <data>
                        <value>
                            <array>
                                <data>
                                    <value>TCPROS</value>
                                </data>
                            </array>
                        </value>
                    </data>
                </array>
            </value>
        </param>
    </params>
</methodCall>

response报文

<?xml version="1.0"?>
<methodResponse>
    <params>
        <param>
            <value>
                <array>
                    <data>
                        <value>
                            <i4>1</i4>
                        </value>
                        <value></value>
                        <value>
                            <array>
                                <data>
                                    <value>TCPROS</value>
                                    <value>sherlock</value>
                                    <value>
                                        <i4>33173</i4>
                                    </value>
                                </data>
                            </array>
                        </value>
                    </data>
                </array>
            </value>
        </param>
    </params>
</methodResponse>

根据协议,调用过程如下:

requestTopic("/test_sub", "/ros_message", ["TCPROS"])

告诉publisher,subscriber准备好了,可以发数据了

requestTopic

_remap_table['requestTopic'] = [0] # remap topic 
@apivalidate([], (is_topic('topic'), non_empty('protocols')))
def requestTopic(self, caller_id, topic, protocols):
    """
    Publisher node API method called by a subscriber node.

    Request that source allocate a channel for communication. Subscriber provides
    a list of desired protocols for communication. Publisher returns the
    selected protocol along with any additional params required for
    establishing connection. For example, for a TCP/IP-based connection,
    the source node may return a port number of TCP/IP server. 
    @param caller_id str: ROS caller id    
    @type  caller_id: str
    @param topic: topic name
    @type  topic: str
    @param protocols: list of desired
     protocols for communication in order of preference. Each
     protocol is a list of the form [ProtocolName,
     ProtocolParam1, ProtocolParam2...N]
    @type  protocols: [[str, XmlRpcLegalValue*]]
    @return: [code, msg, protocolParams]. protocolParams may be an
    empty list if there are no compatible protocols.
    @rtype: [int, str, [str, XmlRpcLegalValue*]]
    """
    if not get_topic_manager().has_publication(topic):
        return -1, "Not a publisher of [%s]"%topic, []
    for protocol in protocols: #simple for now: select first implementation 
        protocol_id = protocol[0]
        for h in self.protocol_handlers:
            if h.supports(protocol_id):
                _logger.debug("requestTopic[%s]: choosing protocol %s", topic, protocol_id)
                return h.init_publisher(topic, protocol)
    return 0, "no supported protocol implementations", []

init_publisher 代码

def init_publisher(self, resolved_name, protocol):
    """
    Initialize this node to receive an inbound TCP connection,
    i.e. startup a TCP server if one is not already running.
    
    @param resolved_name: topic name
    @type  resolved__name: str
    
    @param protocol: negotiated protocol
      parameters. protocol[0] must be the string 'TCPROS'
    @type  protocol: [str, value*]
    @return: (code, msg, [TCPROS, addr, port])
    @rtype: (int, str, list)
    """
    if protocol[0] != TCPROS:
        return 0, "Internal error: protocol does not match TCPROS: %s"%protocol, []
    start_tcpros_server()
    addr, port = get_tcpros_server_address()
    return 1, "ready on %s:%s"%(addr, port), [TCPROS, addr, port]

publisher 检查,是否支持指定协议,如果不支持,则返回1,否则返回0

返回值的第二个参数有三个值,分别是协议类型、ip地址 和 端口

5、unregisterSubscriber

request报文body

<?xml version="1.0"?>
<methodCall>
    <methodName>unregisterSubscriber</methodName>
    <params>
        <param>
            <value>/test_sub</value>
        </param>
        <param>
            <value>/ros_message</value>
        </param>
        <param>
            <value>http://192.168.1.150:43597</value>
        </param>
    </params>
</methodCall>

response报文body

<?xml version='1.0'?>
<methodResponse>
    <params>
        <param>
            <value>
                <array>
                    <data>
                        <value>
                            <int>1</int>
                        </value>
                        <value>
                            <string>Unregistered [/test_sub] as provider of [/ros_message]</string>
                        </value>
                        <value>
                            <int>1</int>
                        </value>
                    </data>
                </array>
            </value>
        </param>
    </params>
</methodResponse>

unregisterSubscriber

@apivalidate(0, (is_topic('topic'), is_api('caller_api')))
def unregisterSubscriber(self, caller_id, topic, caller_api):
    """
    Unregister the caller as a subscriber of the topic.
    @param caller_id: ROS caller id
    @type  caller_id: str
    @param topic: Fully-qualified name of topic to unregister.
    @type  topic: str
    @param caller_api: API URI of service to unregister. Unregistration will only occur if current
       registration matches.    
    @type  caller_api: str
    @return: (code, statusMessage, numUnsubscribed). 
      If numUnsubscribed is zero it means that the caller was not registered as a subscriber.
      The call still succeeds as the intended final state is reached.
    @rtype: (int, str, int)
    """
    try:
        self.ps_lock.acquire()
        retval = self.reg_manager.unregister_subscriber(topic, caller_id, caller_api)
        mloginfo("-SUB [%s] %s %s",topic, caller_id, caller_api)
        return retval
    finally:
        self.ps_lock.release()

取消订阅,固定返回1

6、TCP数据私有协议

首先保证主从机的数据类型一致,包括字段的顺序,实际ROS框架内是通过md5检测,保证数据类型一致的

数据传输前提:

  1. 数据类型一致
  2. 字段名一致
  3. 字段顺序一致

数据传输模式:小端hex

数据包结构

数据域长度 数据域
4 byte n byte

数据域长度固定4byte,长度不包括自身

数据域

数据域根据字段类型解析,ros 通信的内置数据类型有:

原始类型 字节数
bool 1
int8 1
uint8 1
int16 2
uint16 2
int32 4
uint32 4
int64 8
uint64 8
float32 4
float64 8
string n(n > 4)
time 8
duration 8
数组 n(n > 4)

其中,除 string、time、duration数组 类型外的其余类型,直接根据字节数读取即可

string

字符串类型,也可认为是字符数组(则可以和数组类型复用),因为是不定长度,所以需要知道字符串的长度,ROS中使用uint32类型表示长度/数组元素数量,即4byte

所以,如果出现字符串类型,则数据域为:

字符串长度 字符
4 byte n byte

数组

数组类型,因为是不定长度,所以需要知道数组的元素数量,和string同理,ROS 中使用uint32类型表示数组的元素数量,再结合数组元素的类型,即可得到总长度

所以,出现数组类型,则数据域为:

数组元素数量 数组数据
4 byte n byte

如果数组类型是 int32,则数组数据占 4 * n byte,其余类型以此类推

time

ROS 中把 time 单独提取作为基本数据类型,对应 ROS 中的 ros::Time 类,因为我们可以认为是嵌套类型

ros::Time 有两个字段:

  1. sec: uint32
  2. nsec: uint32

所以,time 类型在数据域占8byte,如果出现 time 类型,则数据域为:

sec nsec
4 byte 4 byte

duration

duration 类型和 time 相同,在 ROS 中对应 ros::Duration 类,可以认为是嵌套类型

ros::Duration 有两个字段:

  1. sec: uint32
  2. nsec: uint32

所以,duration 类型在数据域中占8byte,如果出现 duration 类型,则数据域为:

sec nsec
4 byte 4 byte

嵌套类型

嵌套类型可以认为是数据域的组合,如果发现字段类型不是内置数据类型,则可认为是嵌套类型,嵌套类型按照类型的字段,递归处理即可

协议分析示例

示例1:

.msg 文件为:

int8 shutdown_time
string text

主机发出数据为:

shutdown_time = 123
text = abc

从机收到数据为:

08 00 00 00 7b 03 00 00 00 61 62 63

分析如下:

  1. 包头4 byte表示数据与长度

    08 00 00 00,表示数据域长度为8,即后续数据总长度为8

  2. 字段1为shutdown_time,类型是int8,1byte

    7b转10进制,为123

  3. 字段2为text,类型是字符串 (4+n)byte

    4byte 长度:03 00 00 00,表示字符串长度为3,后面3byte 为字符串内容:61 62 63,ASCII转换为:abc)

示例2:

.msg 文件为:

Header header
int8 shutdown_time
int32 shutdown_time2
string text
float32 num
string text2
int8[] data
int16[] data2

Header的数据类型为:

uint32 seq
time stamp
string frame_id

主机发出数据为:

//header一般由ROS系统自己处理,这里写出来是为了方便观察
header.seq = 29;
header.time.sec = 0;
header.time.nsec = 0;
header.frame_id = "";
shutdown_time = 123;
shutdown_time2 = 987654;
text2 = "lmn";
text = "abc";
num = 23.4;
data = [1, 2, 4, 89];
data2 = [11, 22, 908]

从机收到的数据为:

39 00 00 00 1d 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 7b 06 12 0f 00 03 00 00 00 61 62 63 33 33 bb 41 03 00 00 00 6c 6d 6e 04 00 00 00 01 02 04 59 03 00 00 00 0b 00 16 00 8c 03

分析如下:

  1. 包头4byte表示数据域长度

    0x39 00 00 00,10进制57,表明后续数据域长度57byte

  2. 字段1,Header 类型,可以认为是嵌套类型,Header字段如下:

    1. 字段1,seq,uint32,4byte,数据为,0x1d 00 00 00,十进制29;
    2. 字段2,time类型,可以认为是嵌套类型,字段如下:
      1. 字段1,sec,uint32,4byte,数据为:0x00 00 00 00,十进制0;
      2. 字段2,nsec,uint32,4byte,数据为:0x00 00 00 00,十进制0;
    3. 字段3,frame_id,字符串类型,4byte 表示长度,00 00 00 00,表示长度为0,字符串为空
  3. 字段2,shutdown_time,int8,1byte,数据为:0x7b,十进制123;

  4. 字段3,shutdown_time2,int32,4byte,数据为:0x06 12 0f 00,十进制:987654;

  5. 字段4,text,字符串:

    1. 4byte 长度,数据为:0x03 00 00 00 ,表示字符产长度为3;
    2. 字符串内容,数据为:0x61 62 63 ,ASCII对应:abc;
  6. 字段5,num,flota32,4byte,数据为:33 33 bb 41,十进制:23.4;

  7. 字段6:text2,字符串:

    1. 4byte长度,数据为:0x03 00 00 00,表示字符串长度为3;
    2. 字符产内容,数据为:0x6c 6d 6e,ASCII对应lmn;
  8. 字段7,data,int8数组:

    1. 4byte表示数组元素数量,数据为:0x04 00 00 00,表示有4个int8元素:
    2. 数组内容:[0x01, 0x02, 0x04, 0x59,],表示:[1,2,4,89];
  9. 字段8,data2,int16数组:

    1. 4byte表示长度,数据为:0x03 00 00 00,表示有3个int16数据;
    2. 数组内容:[0x0b00, 0x1600, 0x8c03],表示:[11, 22, 908]

7、小结

宗上,整体的从机订阅时序图如下:

ROS应用层通信协议解析