Twin-Timescale Artificial Intelligence Aided Mobility-Aware Edge Caching and Computing in Vehicular Networks

Abstract

In this paper, we propose a joint communication, caching and computing strategy for achieving cost efficiency in vehicular networks. In particular, the resource allocation policy is specifically designed by considering the vehicle's mobility and the hard service deadline constraint. An artificial intelligencebased multi-timescale framework is proposed for tackling these challenges. To mitigate the complexity associated with this large action and search space in the sophisticated multi-timescale framework considered, we propose to maximize a carefully constructed mobility-aware reward function using the classic particle swarm optimization scheme at the associated large timescale level, while we employ deep reinforcement learning at the small timescale level of our sophisticated twin-timescale solution. Numerical results are presented to illustrate the theoretical findings and to quantify the performance gains attained.

Computation Offloading for Vehicular Environments: A Survey

“…The edge can contain one or more servers or even mini data centers to serve clients of the most varied types. As seen in the Figure 4(c), most works have chosen to offload to the edge servers such as [58], [47], and [115].…”

“…Among the selected works, a great majority employ metaheuristics for solving optimization problems. The highlights are particle swarm optimization (PSO) [114], [115], [128], ant colony optimization (ACO) [75], [92], and genetic algorithm (GA) [77], [102], [149], [151]. Other examples of metaheuristics are bat algorithm [83] and iterated local search (ILS) [154].…”

Edge Cache-Assisted Secure Low-Latency Millimeter-Wave Transmission

“…In general, cache status b f,u or cache placement depends on several factors, such as user behavior, information popularity distribution and so on, and it can be decided by several advanced cache schemes, e.g., artificial intelligencebased multi-timescale framework method [32] and deep Qlearning method [37]. Here, we assume that the cache status b f,u has been fixed according to a certain cache strategy, and the required files by the actuators, i.e., c k, f are also given in advance [19], [21].…”

Coordinated Hybrid Precoding Design in Millimeter Wave Fog-RAN

“…where θ (i) l,k and β (i) l,k , respectively, are the value of θ l,k and β l,k at the ith iteration. To this end, (12b) can be transformed into the following convex constraint The convex solvers (e.g., CVX) can be used to solve (16). Summarily, solving the original problem (12), we need to iteratively solve the optimal values of {v l,k }, {β l,k }, {θ l,k }, {τ l,k } via (17).…”

“…where g l ∈ C 1×S , w l ∈ C S ×1 and x CP l , respectively, mean the channel coefficient, precoding vector and transmit signal from the CP to the lth FBSH. Then, the achievable rate of the lth FBSH can be calculated by R CP l = log 2 (1 + γ CP l ), where γ CP l is 2 In general, how to cache contents depends on the user behavior, the delay requirement and the long-term information popularity distribution, which can be realized by learning-based methods, such as based on deep Q-learning [15] and artificial intelligence-based multi-timescale framework [16]. In this paper, we mainly focus on the resource allocation for a given cashing placement.…”