refactored codebase

msg/Metric.msg

-string  component # A metric is affiliated with a roboter component.
+string  domain # The domain the metric
+               # is affiliated with.
 string  label  # Label of the metric.
 string  value  # Value of the metric.
 string  unit   # Unit of the metric.
-float32 critical  # Critical level of the metric.
\ No newline at end of file
+float32 error  # Error level of the
+               # metric.
\ No newline at end of file

msg/State.msg

-string  component # Name of the component, whose
-                  # critical level is monitored.
-float32 critical  # Critical level of the family.
\ No newline at end of file

msg/States.msg

-State[] states # Array of states.

nodes/monitor_distance → nodes/monitor_distance.py

 #!/usr/bin/env pyhton
 
 import rospy
-from monitoring.monitor import Monitor
 from sensor_msgs.msg import LaserScan
+from monitoring.monitor import Monitor
 
 
-def scan_callback(msg):
-    global range_ahead
-    range_ahead = msg.ranges[int(len(msg.ranges) / 2)]
+def calculate_distance(msg):
+    global distance
+    distance = msg.ranges[int(len(msg.ranges) / 2)]
 
 
 if __name__ == '__main__':
     try:
-        range_ahead = 10
-        scan_sub = rospy.Subscriber("scan", LaserScan, scan_callback)
+        sub = rospy.Subscriber("scan", LaserScan, calculate_distance)
         rospy.init_node("monitor_distance")
-        monitor = Monitor(1, 1)
+
+        monitor = Monitor()
 
         rate = rospy.Rate(10)
         while not rospy.is_shutdown():
-            print(range_ahead)
-            if range_ahead > 0.8:
-                critical = 0
+
+            if distance > 0.8:
+                error = 0
             else:
-                if range_ahead > 0.35:
-                    critical = 0.5
+                if distance > 0.35:
+                    error = 0.5
                 else:
-                    critical = 1.0
+                    error = 1.0
 
-            monitor.update_metric("scan", "distance", range_ahead, "unit", critical)
+            monitor.update_metric("kinect", "distance", distance, "meter", error)
             rate.sleep()
+
     except rospy.ROSInterruptException:
         pass

src/monitoring/monitor.py

@@ -2,21 +2,36 @@
 
 import rospy
 import socket
-from monitoring.msg import Monitoring
+
 from monitoring.msg import Metric
+from monitoring.msg import Monitoring
 
 
 class Monitor:
-    def __init__(self, mode=1, freq=1.0):
+    """
+    Each metric should have its own monitor.
+    A monitor has four modes:
+
+    - Mode 1: Publish the last received value.
+    - Mode 2: Publish the lowest received value.
+    - Mode 3: Publish the highest received value.
+    - Mode 4: Publish the average received value
+              over five or more seconds.
+
+    Furthermore, a monitor publishes messages with a
+    certain frequency.
+    """
+    def __init__(self, mon_mode=1, mon_freq=1.0):
         self.log = {}  # Logging dictionary. Is used to accomplish different publishing modes.
-        self.pub = rospy.Publisher('monitoring', Monitoring, queue_size=1)
-        self.mode = mode
+        self.pub = rospy.Publisher('/monitoring', Monitoring, queue_size=1)
+        self.mon_mode = mon_mode
 
-        if not freq > 0.0:
-            rospy.logwarn("Frequency must be greater then 0! Using 1 as frequency!")
-            freq = 1.0
+        if not mon_freq > 0.0:
+            rospy.logwarn("The frequency at which a monitor publishes messages must be greater then 0! "
+                          "Using 1 as frequency!")
+            mon_freq = 1.0
 
-        self.timer = rospy.Timer(rospy.Duration(1.0 / freq), self.publish_monitoring)
+        self.timer = rospy.Timer(rospy.Duration(1.0 / mon_freq), self.publish_monitoring)
 
         self.monitoring = Monitoring()
         self.monitoring.origin = socket.gethostname() + rospy.get_name()
@@ -25,14 +40,14 @@ class Monitor:
     def publish_monitoring(self, event):
         self.pub.publish(self.monitoring)
 
-    def update_metric(self, component, label, value, unit, critical):
+    def update_metric(self, domain, label, value, unit, error):
         def set_metric():
             metric = Metric()
-            metric.component = str(component)
+            metric.domain = str(domain)
             metric.label = str(label)
             metric.value = str(value)
             metric.unit = str(unit)
-            metric.critical = critical
+            metric.error = error
 
             self.monitoring.metric = metric
 
@@ -41,17 +56,17 @@ class Monitor:
 
         if " " in label:  # A label cannot contain whitespace!
             rospy.logwarn("Whitespaces are not allowed in metric labels!")
-        elif critical > 1 or critical < 0:
-            rospy.logwarn("Critical level must be between 1 and 0!")
+        elif error > 1 or error < 0:
+            rospy.logwarn("Error level must be between 1 and 0!")
         else:
             if label in self.log:
                 # Mode 1: Set the last obtained value.
-                if self.mode == 1:
+                if self.mon_mode == 1:
                     set_metric()
 
                 # Mode 2: Remember and set the minimum value,
                 #         obtained after mode selection.
-                elif self.mode == 2:
+                elif self.mon_mode == 2:
                     if value < self.log[label]['val']:
                         self.log[label]['val'] = value
                     else:
@@ -61,7 +76,7 @@ class Monitor:
 
                 # Mode 3: Remember and set the maximum value,
                 #         obtained after mode selection.
-                elif self.mode == 3:
+                elif self.mon_mode == 3:
                     if value > self.log[label]['val']:
                         self.log[label]['val'] = value
                     else:
@@ -71,7 +86,7 @@ class Monitor:
 
                 # Mode 4: Set the average obtained value over five
                 #         or more seconds.
-                elif self.mode == 4:
+                elif self.mon_mode == 4:
                     duration = rospy.get_rostime() - self.log[label]['dur']
                     if duration < rospy.Duration(5):
                         self.log[label]['num'] += 1

src/monitoring/watchdog.py

 #!/usr/bin/env python
 
 import rospy
+
 from monitoring.msg import Monitoring
-from monitoring.msg import State
-from monitoring.msg import States
 
 
 class Watchdog:
-    def __init__(self, mode=1, freq=1.0):
-        self.mode = mode
-        self.pub = rospy.Publisher('states', States, queue_size=1)
-        self.component_states = {}
-        self.observing = {}
+    def __init__(self):
+        self.sub = rospy.Subscriber("/monitoring", Monitoring, self.update_log)
+
+        # Stores the most current values of each metric.
+        self.log = {}
+
+        # Strategies that specify how the critical level of each metric should be aggregated in a domain.
+        # It is possible to specify a different aggregation strategy for every single domain.
+        self.aggregation_strategy_metrics = rospy.get_param("/aggregation_strategy_metrics")
+        # Dictionary that stores the aggregated critical levels of the metrics of a domain.
+        self.aggregated_metrics = {}
+
+        # Strategy that specifies how the critical levels of all domains should be aggregated.
+        self.aggregation_strategy_domains = rospy.get_param("/aggregation_strategy_domains")
+        # Variable that stores the aggregated critical level of all domains.
+        self.aggregated_domains = None
 
-        if not freq > 0.0:
-            rospy.logwarn("Frequency must be greater then 0! Using 1 as frequency!")
-            freq = 1.0
+        # Frequency at which the metrics in the log are aggregated.
+        wdg_freq = rospy.get_param("/wdg_freq", 1.0)
 
-        self.timer = rospy.Timer(rospy.Duration(1.0 / freq), self.update_states)
+        if not wdg_freq > 0.0:
+            rospy.logwarn(
+                "Watchdog: The frequency at which the metrics of the robot are aggregated must be greater then 0! "
+                "Using 1 as frequency!")
+            wdg_freq = 1.0
 
-        rospy.Subscriber("monitoring", Monitoring, self.logging)
+        self.timer = rospy.Timer(rospy.Duration(1.0 / wdg_freq), self.update_condition)
 
     # TODO: Think about efficiency!
-    def logging(self, monitoring):
+    def update_log(self, monitoring):
         origin = monitoring.origin
         metric = monitoring.metric
 
-        component = metric.component
+        domain = metric.domain
         label = metric.label
         value = metric.value
         unit = metric.unit
-        critical = metric.critical
+        error = metric.error
+
+        if domain in self.log:
+            if origin in self.log[domain]:
+                self.log[domain][origin].update({label: {"value": value, "unit": unit, "error": error}})
+            else:
+                self.log[domain][origin] = {}
+                self.log[domain][origin][label] = {"value": value, "unit": unit, "error": error}
+        else:
+            self.log[domain] = {}
+            self.log[domain][origin] = {}
+            self.log[domain][origin][label] = {"value": value, "unit": unit, "error": error}
+
+    def update_condition(self, event):
+        self.aggregate_metrics()
+        self.aggregate_domains()
+
+    # TODO: Add more modes!
+    def aggregate_metrics(self):
+        for domain in self.log:
+            if domain in self.aggregation_strategy_metrics:
+
+                # Strategy 1: Take the highest error level of any metric in a domain.
+                if self.aggregation_strategy_metrics[domain] == 1:
+                    highest_error = 0.0
+
+                    for origin in self.log[domain]:
+                        for label in self.log[domain][origin]:
+                            local_error = self.log[domain][origin][label]["error"]
+                            if local_error > highest_error:
+                                highest_error = local_error
+
+                    self.aggregated_metrics.update({domain: highest_error})
+
+                # Strategy 2: Take the lowest error level of any metric in a domain.
+                if self.aggregation_strategy_metrics[domain] == 2:
+                    lowest_error = 1.0
+
+                    for origin in self.log[domain]:
+                        for label in self.log[domain][origin]:
+                            local_error = self.log[domain][origin][label]["error"]
+                            if local_error < lowest_error:
+                                lowest_error = local_error
+
+                    self.aggregated_metrics.update({domain: lowest_error})
 
-        if component in self.observing:
-            if origin in self.observing[component]:
-                self.observing[component][origin].update({label: {"value": value, "unit": unit, "critical": critical}})
                 else:
-                self.observing[component][origin] = {}
-                self.observing[component][origin][label] = {"value": value, "unit": unit, "critical": critical}
+                    rospy.logwarn("Watchdog: The specified method for aggregating metrics is not recognised!")
+
+            # Default strategy: Take the average error level of every metric in a domain.
             else:
-            self.observing[component] = {}
-            self.observing[component][origin] = {}
-            self.observing[component][origin][label] = {"value": value, "unit": unit, "critical": critical}
+                count = 0.001
+                aggregated_error = 0.0
+
+                for origin in self.log[domain]:
+                    for label in self.log[domain][origin]:
+                        count = count + 1
+                        aggregated_error = aggregated_error + self.log[domain][origin][label]["error"]
+
+                quotient = aggregated_error / count
+                self.aggregated_metrics.update({domain: quotient})
 
     # TODO: Add more modes!
-    def update_states(self, event):
-        for component in self.observing:
-            # Mode 1: Take the average critical level for a component
-            if self.mode == 1:
-                aggregation = 0
-                number = 0
-
-                for origin in self.observing[component]:
-                    for label in self.observing[component][origin]:
-                        number = number + 1
-                        aggregation = aggregation + self.observing[component][origin][label]["critical"]
-
-                quotient = aggregation / number
-                self.component_states.update({component: quotient})
-
-        self.publish_states()
-
-    # TODO: Update names!
-    def publish_states(self):
-        topic = States()
-        for component in self.component_states:
-            state = State()
-            state.component = component
-            state.critical = self.component_states[component]
-
-            topic.states.append(state)
-
-        self.pub.publish(topic)
+    def aggregate_domains(self):
+        # Strategy 1: Take the highest error level of any domain.
+        if self.aggregation_strategy_domains == 1:
+            highest_error = 0.0
+
+            for domain in self.aggregated_metrics:
+                local_error = self.aggregated_metrics[domain]
+                if local_error > highest_error:
+                    highest_error = local_error
+
+            self.aggregated_domains = highest_error
+
+        # Strategy 2: Take the lowes error level of any domain.
+        elif self.aggregation_strategy_domains == 2:
+            lowest_error = 1.0
+
+            for domain in self.aggregated_metrics:
+                local_error = self.aggregated_metrics[domain]
+                if local_error < lowest_error:
+                    lowest_error = local_error
+
+            self.aggregated_domains = lowest_error
+
+        # Default strategy: Take the average error level of every domain.
+        else:
+            count = 0.001
+            aggregated_error = 0.0
+
+            for domain in self.aggregated_metrics:
+                count = count + 1
+                aggregated_error = aggregated_error + self.aggregated_metrics[domain]
+
+            quotient = aggregated_error / count
+            self.aggregated_domains = quotient