Upload New File

helpers/mappings.py

+from abc import ABC
+import tensorflow as tf
+import numpy as np
+
+class Mapping(ABC):
+    @tf.function
+    def map(x):
+        pass
+
+class Active_UEs_from_Connected(Mapping):
+    def __init__(self, demand = "medium"):
+        #Linear mapping from monitoring data with different scaling factors depending on demand
+        if demand == "high":
+            self.k = 0.032709253358325543
+        elif demand == "medium":
+            self.k =  0.018575757575757582
+        elif demand == "low":
+            self.k =  0.007832927210246799
+
+    @tf.function
+    def map(self, x):
+        return tf.cast(self.k * x, tf.float32)
+
+class Cell_Load_from_Active_UEs(Mapping):
+    def __init__(self):
+        #Exponential fit from monitoring data
+        self._w_0 = 0.92972595
+        self._w_1 = 2.25275329
+
+    @tf.function
+    def map(self, x):
+        return tf.cast(self._w_0 * (1 - tf.exp(-(x) / self._w_1)), tf.float32)
+
+class Spectral_Efficiency_from_CQI(Mapping):
+    def __init__(self, T=1/10):
+        #Temperature for the sigmoid function to control smoothness
+        self._T = T
+
+    @tf.function
+    def map(self, x):
+        efficiency = np.array(
+            [0.1523, 0.377 , 0.877 , 1.4766, 1.9141, 2.4063, 2.7305, 3.3223, 3.9023, 4.5234, 5.1152, 5.5547, 6.2266, 6.9141, 7.4063]
+        )
+
+        mapping_soft = 0
+        current_sum = 0
+        for ankerpoint in range(1, 16):
+            spec_eff = efficiency[ankerpoint-1]
+            delta = spec_eff - current_sum
+            mapping_soft += delta * (tf.math.sigmoid((x - ((ankerpoint-1) + 0.5))/self._T))
+
+            current_sum += delta
+
+        return tf.cast(mapping_soft, tf.float32)
+
+class CQI_from_SINR(Mapping):
+    def __init__(self):
+        #Linear fit with softplus squasing from MDT data
+        self._w_0 = 0.41238472
+        self._w_1 = 7.73600132
+
+        lambda x: 15 -  tf.math.softplus(
+            -(tf.math.softplus(0.41238472 * x + 7.73600132) - 15)
+        )
+
+    @tf.function
+    def map(self, x):
+        return  tf.cast(15 -  tf.math.softplus(
+            -(tf.math.softplus(self._w_0 * x + self._w_1) - 15)
+        ), tf.float32)