The invention relates to a distributed
reinforcement learning internet-of-vehicles
resource allocation method based on a multi-round
random search strategy, belongs to the technical field of internet-of-vehicles
wireless communication, and aims to solve the problems that in the prior art, internet-of-vehicles
resource allocation is high in action space complexity, not ideal in
resource allocation optimization and high in resource allocation efficiency. The
successful transmission rate of safe data in the V2V link cannot be ensured; and the
throughput of the V2I link cannot be maximized. According to the method, the power and channel allocation neural network is constructed, the global scheme calculation reward is generated, the experience
pool is optimized by using multiple rounds of
random search, the network is trained, and finally the test and evaluation are performed, so that the resource allocation efficiency and the
network performance are improved. According to the method, the resource allocation action space of
the Internet of Vehicles is simplified, the solving difficulty is reduced, the
algorithm convergence is accelerated through multiple rounds of
random search, the channel allocation performance is improved through
cross entropy loss training, the secure
data transmission of the V2V link is ensured, the
throughput of the V2I link is maximized, and the operation efficiency and the user experience of
the Internet of Vehicles are improved.