Research On UAV Path Planning And Task Offloading Methods Based On Computing Task Validity

The widespread application of Io T technology has promoted the development of terminal intelligence.Deployed in areas such as factories,farms,open pit mines and disaster sites,the terminals can collect environmental information and transform the information into computing tasks.The terminals can compute these tasks locally,or offload them to the cloud for computation.The former usually consumes large amounts of energy,while the latter leads to network congestion.Usually,terminals are limited by their own size and carry limited battery energy,but replacing batteries for them timely is impractical in the above scenario.In addition,the tasks generated by terminals have a valid period,i.e.,terminals generate tasks in real time and need to finish processing these tasks within a valid period,otherwise them will be discarded.A large number of tasks discarded can lead to the system’s inability to make accurate analysis and judgment on the environmental changes in the target area.Therefore,the resource scheduling problem of task validity needs to be addressed to minimize task loss in the presence of terminal energy constraints.The development of UAV technology provides ideas to solve the above problems.On the one hand,UAVs can carry edge computing modules and fly over the target area to provide mobile edge computing services for terminals,which alleviates the problem of ground terminal computing resource scarcity.On the other hand,the terminal working time can be effectively extended by installing an energy transmitting module on UAVs to provide wireless charging service for terminals on the ground.However,under the scenario of considering task validity,how to effectively implement UAV-assisted edge computing and wireless charging also faces various challenges,such as the flight trajectory optimization problem of UAVs and the association strategy problem of terminals and UAVs.Based on the above analysis,this thesis achieves the goal of minimizing the system task loss rate by finishing processing as many tasks as possible under the task validity constraint,and the specific research work is as follows.First,considering an open and unobstructed area(e.g.,farms),where terminals on the ground can only transmit tasks to the computing module carried by UAVs for processing via a wireless channel due to energy constraints.A validity model and caching model for tasks generated by terminals are developed.Considering the association strategy problem of terminals and UAVs and the problem of task offloading in the terminal,where task offloading is limited by both the wireless transmission capability and the computing capability of the edge computing module,the system task loss rate is minimized by jointly optimizing the UAV flight trajectory and association strategy between UAVs and terminals,while the task loss rate between terminals is balanced and the energy consumption between UAVs is balanced.The problem is a mixed-integer nonlinear programming problem with a continuous time series and is difficult to solve through traditional convex optimization methods.Therefore,a deep deterministic strategy gradient algorithm based on reinforcement learning is proposed to solve the above problem.The effectiveness of the algorithm is proved by a series of experimental comparisons.Then,based on the study of line-of-sight,the obscured non-line-of-sight between terminals and UAVs that may exist in urban and rural areas is considered.The terminals have computing capability,but they are limited by battery energy and cannot process all the tasks.An energy transmitting module is installed on UAVs to provide wireless charging service for terminals on the ground to replenish the energy required for local computing,so as to reduce the task loss.A mixed line-of-sight and non-line-of-sight link model is developed,and a realistic non-linear energy harvesting model is used for terminals.Terminal task processing and task queue updating methods are designed under the situation that task processing is limited by both energy and computing capability.Considering computing task validity,the system task loss rate is minimized by optimizing the flight trajectory of UAVs,while the task loss rate of each terminal is balanced and the energy consumption of each UAV is balanced.The problem is an NPHard problem under continuous time series,which is difficult to solve through traditional optimization algorithms.A deep reinforcement learning algorithm is proposed to solve the problem.Simulation results validate the effectiveness of the proposed algorithm.