Research On The Optimal Path Planning Mechanism And Method For Intelligent Vehicle

Intelligent vehicle is a kind of mobile robot that can independently complete a given task in a complex environment without relying on human beings.The optimal path is a main research topic in intelligent vehicles and intelligent transportation systems,and it mainly solves the problem of searching collision-free paths that meet the optimal parameter conditions for intelligent vehicles.The references parameters for the optimal path selection of intelligent vehicle include the shortest path problem,the minimum total cost problem,the shortest travel time problem,and so on,targeted path optimization can be achieved according to different technologies.Quickly and effectively path planning is the primary goal to ensure that the vehicles reach their destination safely,and it is also an important part of the intelligent vehicle system.However,increasingly complex transportation networks have put forward higher requirements for the application and popularization of intelligent vehicles.Therefore,this dissertation did detailed analysis and research on the optimal path selection strategy and algorithm for intelligent vehicle based on the study of artificial intelligence technologies such as reinforcement learning(RL)technology,fuzzy neural network(FNN)technology,and the learning of shortest path algorithms such as A* algorithm.Different optimization strategies are used to solve the problems of path planning algorithm in the field of intelligent driving,such as large calculation amount,poor convergence,complex structure,poor practicability and scalability,that is,the topic of this dissertation is “Research on the Optimal Path Planning Mechanism and Method for Intelligent Vehicle”.The main research contents and innovations of this dissertation are as follows:1)Aiming at the lack of a comprehensive evaluation system in the current research on intelligent vehicle path planning,an optimal path selection parameter system is designed to help the path planning algorithm output the optimal path.Although the traditional A* algorithm can solve the shortest path problem well,the actual transportation network is complex and varied,the path calculated by the shortest path algorithm alone has many inflection points,and the final solution path is usually not the optimal solution.Therefore,we design an optimal path selection strategy based on reinforcement learning and shortest path algorithm.This strategy can help intelligent driving vehicles to adjust and select a collision-free optimal path in real time through interactive information in an unknown environment.Reinforcement learning based on prior knowledge can train the intelligent driving vehicle path planning algorithm,by reducing the calculation time of the state set,the convergence speed of the reinforcement learning algorithm is accelerated,and the cost of invalid sample learning and training is reduced.Then the sensor system of the vehicle continuously interacts with the environment in the unknown environment to obtain feedback,and obtains the corresponding action instructions according to the reinforcement learning algorithm.While avoiding obstacles,according to the selected reference standard,the intelligent vehicle can get the optimal path from the start point to the end point.The search efficiency is improved through the search settings change of the shortest path.This path optimization method can effectively help different types of intelligent driving vehicles to smoothly plan the optimal path in the transportation network under conditions of restricted height,width and weight,as well as accidents and congestion obstacles,and solve the problem of less consideration on the above obstacles in existing researches on intelligent driving vehicles path planning.2)We designed a fuzzy neural network path planning algorithm trained by particle swarm optimization algorithm.The PSO based on the foraging behavior of birds has the advantages of small computational complexity and simple structure.But it is easy to fall into the local extremum problem and fall into an infinite loop.The path planning method combined with the fuzzy neural network algorithm can analyze and distinguish various information in the driving environment,and formulate a specific reaction to guide the intelligent driving vehicle forward,and the neural network has strong fault tolerance,strong self-learning ability and very good adaptability,but there is the problem of too much calculation and slow convergence.In order to solve the problem that the particle swarm algorithm is easy to fall into the local minimum value,an improved particle swarm algorithm is designed that the update method of the inertia weight and the learning factor are optimized.And this algorithm is used to train and optimize the fuzzy neural network weight parameters under the defined rules,to solve the problem of slow convergence of fuzzy neural network algorithm in the application of intelligent driving vehicle path planning.By designing reasonable training rules and fuzzy network structure,the hybrid algorithm can accomplish the path planning mission in a large and complex network.3)The application of artificial intelligence(AI)technology in the field of intelligent driving is very effective.The reinforcement learning algorithm based on the Markov process is one of the main methods of trajectory planning and design in the intelligent decision support technology,but it is extremely difficult to select all hyperparameters at the same time because the reinforcement learning algorithm has huge space of hyperparameters.Therefore,we design to optimize the hyperparameters of reinforcement learning to make it converge quickly and improve the learning efficiency;then the particle swarm optimization algorithm is improved by pre-set setting to reduce the calculation of invalid particles,and increase the efficiency of obtaining the optimal solution.Finally,a hybrid algorithm is used for path planning,the individual optimal particle and global optimal position fitness of the improved particle swarm optimization algorithm are corrected by the correction amount,so as to achieve a simple and efficient path planning that does not fall into a local optimal solution.