Parallel distributed computing system and method for flight control agent

Abstract

The invention discloses a parallel distributed computing system and method for a flight control agent. The system comprises a global scheduling module, a data processing module and an analogue simulation module, and is designed for reinforcement learning in the field of aviation control. The problem that other flight simulators cannot uniformly and directly complete reinforcement learning parallel computing is solved; according to the calculation method, main calculation in the air aviation field can be included and distributed. The problem that an analogue simulation calculation process can achieve efficient parallel processing on a plurality of machines is solved, and the method can also be suitable for large-scale calculation in a cluster; according to the technical scheme, the system is characterized by comprising a deep reinforcement learning part, a global scheduling part and a local scheduling part which are used for task scheduling and process scheduling; a state control program is used for supplying an information state to a global scheduling algorithm; a data synchronization part is used for synchronizing the calculated data. The invention has a high-throughput and low-delay data transmission capability; The dynamic construction of flight training tasks is supported.

Description

technical field [0001] The invention relates to the technical field of intelligent air traffic control, in particular to a parallel distributed computing system and method for flight control agents. Background technique [0002] Free air combat represents the future development direction of drones. Air combat behavior is expressed in the form of maneuvering trajectories, and its purpose is to obtain the advantage of the battlefield situation through maneuvering, constitute the conditions for weapon launch, maximize the performance of weapons, destroy the enemy to the greatest extent, and protect our pilots. [0003] In the new syllabus of free air combat training, the aircraft is required to have higher precision and quicker response to defend against the attack of enemy aircraft and make corresponding attacks. Good mobility. To protect the lives of pilots, unmanned aerial vehicles are playing an increasingly important role in air combat. But keeping pilots safe isn't the...

AI Technical Summary

Problems solved by technology

The deep reinforcement learning process of unmanned combat aircraft needs to feed back and update the parameters of the neural network through random sampling and judging the value of the results, while the air combat process takes place in a three-dimensional space, tracking multiple parameters such as position and speed to the observation space. The size has caused exponential growth, and the existing single-agent training environment cannot afford the research needs of intelligent air combat

Images