Research On Key Technologies Of Autonomous Navigation For Unmanned Vehicles

Unmanned vehicles are the product of the development of computer science,pattern recognition and intelligent control technology.They have very broad development prospects in urban transportation,urban security and military patrols.At present,aiming at alleviating urban traffic pressure and reducing urban traffic accidents,unmanned vehicles used in urban transportation and urban security have become a research hotspot.In the field of urban transportation and urban security,the unmanned vehicle should have the ability to perform tasks independently.Therefore,under the urban road environment,the independent autonomous navigation of the unmanned vehicle is a primary orientation for the development of the autonomous navigation technologies.The autonomous navigation of an unmanned vehicle means that it can reach the designated destination autonomously and safely without human intervention.The main research directions include environmental perception,path planning,decision making control and positioning and navigation,among which environmental perception and path planning are the most basic and most critical issues.Therefore,this paper focuses on these two directions.Environmental perception means that an unmanned vehicle uses a variety of detection devices(such as camera,laser,radar)to sense the environment outside the vehicle.In the environment perception,the unmanned vehicles use the visual sensor detection like camera to realize autonomous navigation is called an unmanned vehicle based on visual navigation.Since the visual detection device is not only inexpensive compared to a detection device such as laser and radar,but also provides a rich external environment information for the unmanned vehicle,which is advantageous for the unmanned vehicle to sense the external environment,the visual navigation-based unmanned vehicle is current hot research content.For unmanned vehicles in urban traffic environment,the main task of vision-based environment perception is to detect and identify surrounding urban road environments,including lane line detection and recognition,road marking detection and recognition,road obstacle detection.Due to the complex traffic environment of urban roads and uncontrollable natural environment,when the detection and recognition algorithm for target detection and recognition,the rate of the target detection and recognition is not high,and the real-time performance of the algorithm is poor.Path planning refers to the unmanned vehicles use the information obtained by the environment perception to plan a reasonable route from the starting point of mission to the ending point.Due to the path planning algorithm has problems such as low search precision and stagnation of search in a specific situation,resulting in low path planning accuracy and poor real-time performance.In view of the above problems,this paper takes visual navigation as the main line.Based on the current research,the lane line detection algorithm,the road marking detection and recognition algorithm,the road obstacle vehicle detection algorithm,the path planning algorithm have been further studied to provide an effective solution for the research of autonomous navigation technologies of unmanned vehicles.The main work and innovations of the thesis include:(1)Lane line detection.The model-based lane line detection algorithm is detailed analysis,including linear model,parabolic model and cubic curve model.Aiming at the shortcomings of the current algorithm,a lane line detection algorithm based on convex curve model is proposed.Firstly,the left and right lane lines are detected by using the convex curve model,and then the results of detection are fitted and reconstructed by the least squares method.The proposed algorithm has highter detection precision and improves the narrow application range,insufficient detection ability and poor robustness of the current algorithm.(2)Arrow mark detection and recognition.The detection and recognition algorithm of road arrow markings based on biologically visual perception model and discriminative model is studied.By analyzing the advantages and disadvantages of the two models,a hybrid algorithm combining two models is proposed.The advantages of each other are used to make up the shortcoming of each other.The proposed algorithm improves the speed and accuracy of the detection and recognition of arrow markings in complex environments.(3)Vehicle detection.This part mainly analyzes and summarizes the vehicle detection algorithm based on the Deformable Part Model(DPM).For the classic DPM detection algorithm,there is a small target vehicle miss detection problem.Aiming at the problem of small target vehicle miss detection problem,this paper replaces the fixed-scale pyramid model in the classical DPM algorithm with the variable-scale pyramid model so the improved DPM vehicle detection algorithm based on adaptive pyramid model is proposed.In order to improve the construction speed of the adaptive pyramid model,the fast pyramid estimation theory is used to construct the adaptive pyramid.The improved algorithm reduces the missed detection rate of small target vehicles and improves the detection ability of road vehicles.(4)Path planning.In order to solve the problem of low accuracy of unmanned vehicle path planning in complex road environment,the path planning based on swarm intelligence algorithm is mainly studied,including chicken swarm optimization algorithm,ant colony optimization algorithm and particle swarm optimization algorithm.And an improved particle swarm optimization algorithm is proposed.By improving the parameters of the classical particle swarm optimization algorithm and adding various updating strategies,the search precision and search stagnation of the particle swarm optimization optimization algorithm is improved and the path planning accuracy of the unmanned vehicle is improved.