Mobile Edge Computing for Traffic Management Mechanism using HAPS (High Altitude Platform Station)

MEC (Mobile Edge Computing) is a new paradigm to accelerate artificial intelligence (AI) applications by leveraging computing resources on the network edge and can be used to improve intelligent transportation systems (ITS). HAPS is deployed in the stratosphere to provide wide coverage and strong computational capabilities. It is suitable to coordinate terrestrial resources and store the fundamental data associated with ITS-based applications. Performing task offloading and data caching at (RSU’s) in a cooperative manner can reduce the heavy backhaul load and the retransmission of content downloading.

In this research smart vehicles, Roadside Units (RSU’s), and HAPS, are integrated to build a computation framework for ITS, where the HAPS & RSU’s data library stores the fundamental data needed for the users (vehicles). We are analyzing, computing offloading policy to minimize low-latency and high-bandwidth services minimize the network cost at the user equipment (UE) side, while satisfying the constraints of task offloading deadline, the cache capacity at HAPs and the computing capability of MEC servers.  

Introduction to the Problem

With the rapid development of the internet of things (IOT), various applications such as security monitoring, online games, face recognitions will put forward more stringent requirements in service delay and energy consumption. At the same time, with the arrival of the sixth-generation communication network (6G), many small roadside stations (RSU’s) are deployed intensively in the cell, which may cause network performance degradation and backhaul congestion. Mobile edge computing (MEC) is a novel computing model that provides users with the required services and computing requirements at the edge of the network. MEC integrates wireless network and internet technologies, which can effectively solve the resource shortage problem. In particular, it meets the key requirements of the next generation internet and future mobile bandwidth scenarios.

Because of the differences of scale, diversity, and timeliness, the MEC server has great difficulty in processing and analyzing these data. Secondly, compared with HAPS and RSU’s, MEC server has limited computation and storage resources. The computation results of data are time lag, and the retention and reuse of these data are often ignored. To address these challenges, in this paper, we are focusing on minimizing the delay of the system by optimizing computation offloading and caching decisions as well as bandwidth and computing resource allocations. The simulation results highlight the benefits of HAPS computing for mitigating delays and the significance of caching at network edges.

Related Work to this Problem

There are some proposed solutions based on an Artificial Neural Network (ANN) that is designed to detect and classify the levels of congestion on roads, as well as suggest new routes for drivers to enable them to avoid congested roads. When a vehicle receives a message about a congested road, it can decide whether to keep to its current route or calculate the advantages of an alternative route. Another proposed solution is based on SCORPION that carries out congestion detection and traffic classification using K-NN (k nearest neighbors) in accordance with the average speed and the density of the path.

Compared with FOREVER, these solutions lack some features. All the vehicles require information about new routes from a central entity (RSU (roadside unit). This often causes a longer processing time for the routes and a high traffic bandwidth in the network helps to solve the problems of the distributed routing system observed in some case studies. Since each FOG has knowledge of a particular region of the map and can thus avoid allocating the same route to several vehicles in its own region.

FOREVER uses RSU and sensors that are distributed in the entry map to make a set of FOG. This requires the proposed solution to divide the city into areas and implies that each area will be responsible for a single FOG, which is independent of each other. A single RSU is included for each FOG and the RSU is homogeneously distributed in the map to achieve full coverage. If the vehicle is at the FOG intersection, the closest RSU is chosen.

FOREVER also uses a virtual section of the map called Area of Knowledge (AoK). This area is a region where each RSU has knowledge of the roads (their features and current situation) and can compute a new route for the vehicles that are inside its RSU. Another parameter used by the system is the route size factor. The route size factor determines how much longer (in percent) the new route may be in relation to the current route. This parameter is set by the user in the vehicle embedded system.

We will first describe system model including network, communication, computing, and caching model. Then we formulate the data and computing offloading problem in details.

Here there are 3 constraints. The task execution delay should not exceed the maximum completion time i.e., the deadline of user's experience if the user n chooses local or edge computing. The computing resources allocated by each MEC server to all users cannot exceed its maximum processing capacity. The optimization objective function of the total delay of all CAVs over several time slots, which consists of the delays under three ways: local, edge, and HAPS computing.

Proposed Optimization Methods and Analysis

The key challenges of optimization problem in equation (5) are that: firstly, the eq (5) is a MINLP problem, which is NP-hard due to its complexity to the discrete binary constraints. Furthermore, because the objective function is a mix of variables it is also tough to solve in polynomial terms. In the following section, we'll use scaling approaches, and the relaxation method to break down the original problem into sub parts for task offloading resource allocation from CAV’s to RSU/RSU/HAPS.

We get the solution of the original problem by solving the convex optimization problem. Through dual decomposition method, we provide a distributed algorithm for optimal resource allocation policy. The solution to the dual problem can be broken down into two parts. The uplink transmission time and the offloading policy are included in Level 1 of the inner layer reduction in the outer layer, level 2, is the maximization of Lagrange multipliers.

We formulated the problem as a delay minimization problem, a mixed integer nonlinear optimization problem where we first focused on optimizing the computation offloading and then focused on the allocations for bandwidth and computing resources. For now, we’re reviewing the optimization methods and implementing the next phases.

LinkedIn Profiles of the Members of Team

Chaitanya Rayapalam: https://www.linkedin.com/in/tanya-rcl/

Gayatri Sravanthi Kuntla: https://www.linkedin.com/in/sravanthi-reddy-7a35ba84/

Shalaka Kulal: https://www.linkedin.com/in/shalaka-kulal-992422101/