Delay-Sensitive Dependent Tasks Offloading and Resource Allocation in VEC: A Deep-Reinforcement Learning Approach

Vehicle Edge Computing (VEC) has become a new computing paradigm in the field of intelligent transportation systems. In the VEC environment, system performance is seriously challenged by delay-sensitive and dependent tasks. To address this issue, this paper investigates the joint problem of task offloading and resource allocation in VEC, taking into account and highlighting the delay sensitivity and dependencies of tasks. The corresponding mixed-integer nonlinear programming problem is formulated. Then this joint optimization problem is modeled as a Markov decision process. In order to solve the problem, an algorithm for joint optimization Delay-Sensitive and Dependent Tasks offloading decision and resource allocation in VEC Based on Double Deep Q-Network (DDT-DDQN) is proposed. Given the resource allocation and bandwidth allocation, the task queues are arranged by a prioritization policy, which takes into account the dependencies of tasks and delay-sensitive of tasks, and the Double Deep Q-Network (DDQN) is used to solve the offloading decision and channel allocation.Then, given the offloading decision and channel allocation schemes, the DDQN is utilized to solve the problem the computational resources and bandwidth allocation. Simulation results show that the DDT-DDQN algorithm proposed in this paper outperforms other comparative algorithms in terms of task completion rate and significantly improves the network performance, given the same task arrival rate, computational resources and taking into account the inter-task dependencies.