Deep reinforcement learning for adaptive network slicing in 5G for intelligent vehicular systems and smart cities
Systems and methods for processing a service request within a network environment can include a first cluster of fog nodes that execute service tasks. The cluster can include a primary fog node and nearest neighbor fog nodes. The primary fog node can receive, from the network, a service request, determine service request resource data that includes a first time, quantity of resource blocks required to serve the request, and a hold time required to serve the request locally. An edge controller, connected to the network and the first cluster, can receive, from the primary fog node, the service request resource data, identify available resources at the nearest neighbor fog nodes and the primary fog node, and determine whether resource blocks are available to fulfill the service request using deep reinforcement learning algorithms. The edge controller can also refer a rejected service request to a cloud computing system for execution.
1. A method for processing a service request within a network environment, the method comprising:
receiving, by an edge controller from a primary fog node in a first cluster of fog nodes, a service request resource data for a service request, wherein the first cluster of fog nodes is configured to execute service tasks received from the network environment, the first cluster of fog nodes further including nearest neighbor fog nodes of the primary fog node;
identifying, by the edge controller and based on the service request resource data, available resources at the nearest neighbor fog nodes and the primary fog node;
determining, by the edge controller, whether the nearest neighbor fog nodes or the primary fog node have resource blocks available to fulfill the service request;
identifying, by the edge controller and using a DRL algorithm, that the service request can be fulfilled in response to determining that (i) the nearest neighbor fog nodes or the primary fog node have the resource blocks available to fulfill the service request and (ii) expected future reward for saving the resource blocks available to fulfill the service request for future service requests is less than an expected reward associated with serving the service request at another of the nearest neighbor fog nodes or the primary fog node; and
serving, by the edge controller, the service request at one of the nearest neighbor fog nodes or the primary fog node based on determining that (i) the one of the nearest neighbor fog nodes or the primary fog node has the resource blocks available to fulfill the service request and (ii) the expected future reward for saving the resource blocks available is less than the expected reward associated with serving the service request at the another of the nearest neighbor fog nodes or the primary fog node.
2. The method of claim 1 , wherein:
the service request is received, by the primary fog node and from the network environment, the service request including a timestamp and a utility, and
the service request resource data includes the timestamp, a quantity of resource blocks required to serve the service request, and a hold time required to serve the service request locally by any one of the nearest neighbor fog nodes or the primary fog node.
3. The method of claim 1 , wherein the edge controller is the primary fog node or one of the nearest neighbor fog nodes in the first cluster of fog nodes.
4. The method of claim 1 , wherein the edge controller is centrally located in a geographic area of the first cluster of fog nodes.
5. The method of claim 1 , further comprising learning, by the edge controller and using the DRL algorithm, how to (i) allocate limited resources in the first cluster of fog nodes for each service request and (ii) maximize grade-of-service (GoS), wherein the GoS is a proportion of a quantity of served high-load service requests to a total number of high-load service requests in the first cluster of fog nodes.
6. The method of claim 1 , further comprising assigning, by the edge controller, a reward for serving the service request.
7. The method of claim 1 , further comprising serving, by the edge controller, the service request based on determining that the expected reward associated with serving the service request exceeds a threshold value.
8. A system for processing a service request within a network environment, the system comprising:
a first cluster of fog nodes configured to execute one or more service tasks that are received from the network environment, wherein the first cluster of fog nodes includes a primary fog node and nearest neighbor fog nodes of the primary fog node, the primary fog node configured to:
receive, from the network environment, a service request, wherein the service request includes a first time and a utility;
determine, based on the service request, service request resource data, wherein the service request resource data includes the first time, a quantity of resource blocks required to serve the service request, and a hold time required to serve the service request locally by any one of the nearest neighbor fog nodes or the primary fog node;
an edge controller communicatively connected to the network environment and the first cluster of fog nodes, the edge controller configured to:
receive, from the primary fog node, the service request resource data;
identify, based on the service request resource data, available resources at the nearest neighbor fog nodes and the primary fog node;
determine whether the nearest neighbor fog nodes or the primary fog node have resource blocks available to fulfill the service request;
identify, using a deep reinforcement learning (DRL) algorithm, that the service request can be fulfilled in response to determining that (i) the nearest neighbor fog nodes or the primary fog node have the resource blocks available to fulfill the service request and (ii) expected future reward for saving the resource blocks available to fulfill the service request for future service requests is less than an expected reward associated with serving the service request at another of the nearest neighbor fog nodes or the primary fog node; and
serve the service request at one of the nearest neighbor fog nodes or the primary fog node based on determining that (i) the one of the nearest neighbor fog nodes or the primary fog node has the resource blocks available to fulfill the service request and (ii) the expected future reward for saving the resource blocks available is less than the expected reward associated with serving the service request at the another of the nearest neighbor fog nodes or the primary fog node.
9. The system of claim 8 , wherein the edge controller is the primary fog node or one of the nearest neighbor fog nodes in the first cluster of fog nodes.
10. The system of claim 8 , wherein the edge controller is centrally located in a geographic area of the first cluster of fog nodes.
11. The system of claim 8 , wherein the primary fog node and the nearest neighbor fog nodes in the first cluster of fog nodes are communicatively connected via optical links.
12. The system of claim 8 , wherein the edge controller is further configured to assign a reward for serving the service request.
13. The system of claim 8 , wherein the edge controller is trained to allocate limited resources in the first cluster of fog nodes using a Deep Q-Network (DQN).
14. The system of claim 8 , wherein the service request includes at least one of (i) providing smart lighting and automating public buildings, (ii) managing air quality and monitoring noise, (iii) determining and providing smart waste management and energy consumption management, (iv) providing smart parking assistance, (v) providing in-vehicle audio and video infotainment, (vi) executing a driver authentication service, (vii) monitoring structural health of buildings, (viii) managing and providing safe share rides, (ix) executing a smart amber alerting system, (x) executing a driver distraction alerting system, and (xi) monitoring autonomous driving, wherein the service request is associated with a geographic area of the network environment.
15. The system of claim 8 , wherein the service request includes one or more of adjusting smart lighting and automating public buildings in a geographic area associated with the network environment, adjusting air quality and monitoring noise in a geographic area associated with the network environment, making determinations about smart waste management and energy consumption management in a geographic area associated with the network environment, smart parking assistance in a geographic area associated with the network environment, audio and video infotainment in vehicles in a geographic area associated with the network environment, authenticating drivers in a geographic area associated with the network environment, monitoring structural health of buildings in a geographic area associated with the network environment, determining and providing safe share rides in a geographic area associated with the network environment, and executing a smart amber alerting system in a geographic area associated with the network environment.
16. The system of claim 8 , wherein the service request includes execution of driver distraction alerting systems in a geographic area associated with the network environment.
17. The system of claim 8 , wherein the service request includes providing and managing autonomous driving in a geographic area associated with the network environment.