Unmanned aerial vehicle cluster routing calculation method based on anti-fact strategy gradient

Abstract

The invention discloses an unmanned aerial vehicle cluster routing calculation method based on an anti-fact strategy gradient, and the method comprises the steps: employing a COMA dynamic adaptive reinforcement learning algorithm, wherein the COMA algorithm employs a "centralized training-distributed execution" hybrid architecture. The method can effectively aim at routing scene with high network node dynamics and large in-network flow volatility. According to the method, the COMA algorithm can effectively balance the network average survival time and the data packet transmission success rate for a routing scene with high network node dynamics and large in-network flow fluctuation, so that an efficient and stable intelligent routing strategy is realized; the routing strategy can be better and dynamically adjusted, and the global optimal response to the network state is realized. According to the design that any one random node in the selected area of the data packet serves as the next hop transmission node, the problem that the action space is large in the multi-agent environment is solved through the design, the stability of the algorithm is improved, and interference caused by node mobility to training is reduced to a certain extent.

Description

technical field [0001] The invention relates to the technical field of unmanned aerial vehicles, in particular to a method for computing unmanned aerial vehicle cluster routing based on counterfactual policy gradients. Background technique [0002] For the network routing problem, its fundamental task is to provide end-to-end service quality assurance to users. In this regard, how to balance the performance and cost of routing methods is the core scientific issue. In addition, the volatility of network traffic and the dynamics of nodes also pose a high challenge to the robustness of routing algorithms. [0003] Therefore, starting from the core idea of ​​the routing method, we will introduce the back pressure routing method and Q-routing routing method most relevant to this solution, including: [0004] 1) The core idea of ​​back pressure routing is to apply the cache queue to the devices in the network, and use the Lyapunovoptimization optimization method to optimally con...

AI Technical Summary

Problems solved by technology

[0008] Considering that the application scenarios of existing intelligent routing methods are mostly fixed static topology structures, efficient routing methods for dynamic network nodes are still to be explored

Images