Background Modeling And Application Using Model-to-Model Distance In Complex Scenes

Autonomous or self-driving vehicles is the inevitable trend of Automobile industry development,but also the heart of academia of artificial intelligence research.Decision making and planning is the important algorithm to ensure safe driving of unmanned vehicles.Its main task is to plan the trajectory of the vehicle in the local perception environment.In general,local planning for autonomous or intelligence driving is divided into two major parts: behavioral planning and trajectory generation.Vehicles must ensure safety and collision-free,to abide by the traffic rules,anthropomorphic and other elements when tracking the generated trajectory.In this paper,a local path planning decision-making algorithm is proposed aimed at the unmanned vehicles navigation in a dynamic,multi-lane environment.The main contribution of our work in this paper as following:How to implement anthropomorphic behavioral decision-making for Autonomous vehicles in high-dynamic multi-lane complex scene? For this question,we present an effective behavioral programming algorithm.In this algorithm,unmanned vehicles adjust their own speed,acceleration real time and dynamically,and choosing the right driving state between follow the front car or lane change.Then determine the local start and goal points to generate the initial trajectory.Firstly,we use PID algorithm to dynamically adjust driving reference speed and acceleration to maintain a safe distance between vehicles.It can lead smooth speed and prevent the large acceleration value driving situation.Then we build lane change status transfer module,and give a detailed analysis of lane change trigger conditions and lane keeping conditions.Finally,after determine the local end and the starting point,we use cubic spline interpolation to fit the initial path.The whole process of behavior planning provides a priori conditions for subsequent path optimization.The final trajectory of the autonomous vehicles must meet a variety of external conditions,such as reference speed,acceleration,vehicle motion constraints,distance to obstacles,etc.For those constraints,this paper presents a trajectory optimization method based on nonlinear optimization.The optimization method adjusts the coordinate information and the attitude information of the path point to get the trajectory which can meet multiple constraints to ensure the safety comfort,anthropomorphic of the vehicle driving process.In this paper,we establish an optimization equation with the square sum of multiple objective functions,related to unmanned vehicle speed,acceleration,motion constraints,and the distance to the obstacle,smoothing of trajectories,etc.We build a graph optimization model,and use The Levenberg-Marquardt optimization algorithm to solve the optimization equation.The final trajectory can reach our expectations.The third work in this paper is simulation system.We do the simulation work based on the Udacity platform,And established a complete multi-lane multi-vehicle driving simulation environment,equipped with the proposed decision making and planning algorithm in this paper.The simulation results show that the simulated vehicle can plan a satisfactory local path under different driving scenarios,which verifies the reliability and effectiveness of the proposed algorithm.