Research On Path Planning For Carrier-based Aircraft Tractor Based On Inverse Reinforcement Learning

Carrier-based aircraft tractor is an important engineering vehicle on the carrier deck,which plays a key role in use aviation fuel effectively.Whether it is transport aircraft on the deck or on the hangar,carrier-based aircraft need cooperation of tractor to finish these mission,which doesn't on the take-off state.The working environment of carrier-based aircraft is complicated.So carrier-based aircraft tractor should be cope with different mission and adapt to different sea condition.Besides,they can draw carrier-based aircraft to the target position smoothly and accurately by themselves.Those complicated environment and mission make carrier-based aircraft tractor different with other engineering vehicle.With the deep research of unmanned vehicles,the driverless technology of carrier-based aircraft tractor could become a future development trend as well,which is of great significance to improve it operation efficiency and ensure the safety of personnel.Therefore,dynamic path planning is the key of driverless technology of carrier-based aircraft tractor when it on the limited space of deck.This article attempts to use inverse reinforcement learning to address the problem of path planning of carrier-based aircraft tractor system,which base on summarize and analyze used method of path planning,and learn the trajectory of expert to obtain the best policy of path.First of all,this article introduces the history of path planning.Given the characteristic of dynamic environment of carrier deck,we use Markov decision processes to build the model of drawing processes.Secondly,we propose a method based on reverse reinforcement learning to address the problem of path planning of carrier-based aircraft tractor.And we establish a set of mechanisms of inverse reinforcement learning,which base on the path planning condition of carrier-based aircraft tractor.Those mechanisms include the characteristic value,driving style and weight value.Thirdly,the dynamic model of carrier-based aircraft tractor system,a virtual sensor model,an obstacle model and carrier deck model would be established in the carrier-based aircraft tractor driving and simulation system,which developed by us.We would experiment inverse reinforcement learning result on this platform.Finally,the tractor should learn and analyze the expert trajectory of different driving style and drawing mission.We experiment on the carrier-based aircraft tractor driving and simulation system to proving the correct and effective of the method.