Research On Resource Allocation Algorithm In MIMO Wireless Communication System

Multiple-Input Multiple-Output(MIMO)system is the most spectrum efficient communication system as we known.The energy harvesting(EH)refers to the communication system has an energy harvesting device,which can harvest energy from renewable energy sources such as solar energy,vibration and wind energy.The combination of EH technology and MIMO system can achieve the goal of energy saving in wireless communication system and solve the problem of spectrum resource shortage.It is one of the development trends of green communication in the future.This paper focuses on the power allocation of the EH-MIMO wireless communication system and the main work is as follows:(1)Due to harvest energy has the characteristics of randomness and suddenness in this system,and the channel is changeable and fading.We cannot obtain the future states of energy arrival and channel.Therefore,traditional offline algorithm based on non-causal information is no longer suitable.In order to improve the capacity and the utilization efficiency of energy harvesting,we study a resource allocation problem in EH-MIMO system by using a table-based reinforcement learning algorithm.Firstly we construct a model-free MDP structure for this multiple-slot?optimization problem.In order to learn the map between environment and agent,we use two table-based reinforcement learning algorithms to get suboptimal transmission policies through training Q-table which solve this optimization problem of maximizing system throughput.Finally,simulation results show that both algorithms can achieve convergence.Meanwhile,average throughput performance of it is second only to the offline algorithm and better than other benchmark policies.(2)The table-based reinforcement learning algorithm obtains the convergent Q-table through continuous interaction between the agent and the environment.Then the agent can get a suboptimal transmission policy according to Q-table.In our scenario,with increase of the number of antennas in our EH-MIMO system,the dimension of state and space increases exponentially,which will occupy an enormous amount of memory in transmitter and deteriorate the performance of?the system.Since algorithm complexity also increases exponentially with a rapid dimension increasing,we need to consider?another problem is “dimension disaster”.In order to overcome this problem,we develop a value function approximation SARSA algorithm.Using the principle of Tile-Coding and the main characteristics of optimization problem in the system,we construct three group base functions.By taking the vector product of the basis function with its weight we can get an approximate action value function.The approximate SARSA algorithm uses the approximate action value function to obtain the mapping relation between state and action,which can find the suboptimal transport strategy.Simulation results show that the approximate SARSA algorithm can also obtain a suboptimal transmission policy.Compared with SARSA,this algorithm has a faster convergence speed.Meanwhile,since it does not need to store a table,it does not occupy too much memory at the transmitter.It is more suitable for the system model with infinite and continuous state space dimension.