| Intelligent algorithm is applied to the field of mobile robot so that it truly becomes independent, with independent positioning, probing paths and planning ability.In the complex obstructions environment, if you just use the classic computer programming calculation algorithm to plan the path, the time complexity is very huge, people solve the same problem, however, is very simple, because the special structure of the human nervous system has the ability to solve complex problems.Taking advantage of mimicking human nervous system neural computation method, the robot path planning problem was resolved and the following work was done.Firstly,for mobile robot path planning positioning,a kind of structure neurons with multilateral method positioning method is proposed.Multilateral method can take advantage of the sensor data, but will inevitably introduce measurement errors, resulting in positioning solvable equation. Structure of neurons form neural networks are used to solve the positioning equation, which overcomes the measurement error, but also to avoid the solution of linear equations in the ill-conditioned matrix problem.The simulation shows that the proposed algorithm can effectively overcome the measurement noise,solve the real robot positioning coordinates.Secondly,for the deficiencies of traditional obstacle avoidance method,a new methed is proposed by using fuzzy control theory combined with BP neural network algorithm for mobile robot obstacle avoidance, and the memory capacity of this algorithm is analyzed.The traditional obstacle avoidance algorithm requires a large storage space for obstacle avoidance rules, while taking advantage of the limited experience of the robot, effects are not good.However, the present algorithm neural network has a huge memory capacity and can be on-line training and added avoidance rules, and the generalization ability of the network makes the control quantity more appropriate.Thirdly,in the path of space exploration algorithm, the environmental obstacles edge with "within the curved arc" fitting method is proposed in this paper.In this paper, the simulation analysis pointed out that "spline interpolation method" and "radial basis function method" in the curve fitting process is easy to put obstacles into the path space.However, the present algorithm with a small amount of wasted space path, the robot can effectively reduce the presence of the obstacle walking space.Fourthly,in the local path planning algorithm study,for many problems in artificial potential field method,a combination of a variety of improved artificial potential field algorithm is proposed.This paper presents a grid method with potential field method to solve that it can not be applied to global path planning;Proposed dynamic target point ways makes the robot out of potential field local minimum points;Chain network, the method is used to prevent shocking;By combining the above algorithm simulation program proves the feasibility of the algorithm. Fifthly,in the global path planning algorithm, an improved growing nerve gas which combine two A*algorithm path planning method is proposed in this paper.Simulation results show that growing neural gas(GNG) algorithm to generate a map mapping is slower and there will be individual error path, in the subsequent path search, one time A*algorithm can’t find the shortest path. Through the GNG algorithm convergence analysis and proof,an improved GNG algorithm is proposed in this paper. Map generation time is effectively reduced..the path search phase using twice A*algorithm is proved that the final path is shorter. In this paper error path in the map mapping is also solved.Simulation results show that the proposed method is effective for path planning. |
PDF Full Text Request