Research On Path Planning Of Mobile Robots Based On Neural Networks

The mobile robot is one of the most important branches in the field of robot. It is widely used in various areas, such as industry, agriculture, military and education. And path planning is one of the most important issues of mobile robot. This thesis concentrates on the path planning of the mobile robot in static environment based on the technique of neural networks.In this thesis, path-planning is described as proposing assistance to robot to achieve the best path from starting point to goal point by avoiding all barriers. In different static environment, a corresponding solution is also suggested.Firstly, to solve the problem of path planning under fully unknown obstacles space, the method is researched based on Elman network. The network is built which the input is the environment information detected by senor, and the output is the corner of mobile robot. The network is getting stability by training at a high speed. The simulation results confirm the approach is feasible.Next, in static certain environment, based on neural networks, the collision punish function is established. An optimization model is put forward, which is composed by the collision punish function and the length of path. And then the model is solved by using simulated annealing algorithm (SAA). It is well known that convergence speeds of classical SAA are slow. So aiming at improving the convergence speeds of SAA, this algorithm is modified. Then a new modified SAA is put forward. It gets faster convergence speeds and realizes global path planning of mobile robot.At last, a method of path planning based on neural network and genetic algorithm is proposed. The neural network model is used for depicting the information of environment around the robot, and then the output of the NN model is used to construct the fitness function of the genetic algorithm, which is used to optimize the path. In the genetic algorithm, the two-dimensional coding for the via-points of path is converted to one-dimensional one. Meanwhile, a genetic simulated annealing algorithm is developed to take place of the genetic algorithm, which is a hybrid of genetic and simulated annealing algorithm. The simulation results show that the genetic simulated annealing algorithm is faster to plan a better path in complex environment than the simple genetic algorithm, and validated the effectiveness of the proposed approach.