Research On Robot Path Planning Based On Bayes Decision And Ant Colony Algorithm In A Complex Environmnet

Mobile robot path planning is an important issue in robot control technology, which has been greatly developed and applied. But in actual applications, the working environments for mobile robot are complex with dangerous traps and obstacles, such as concave obstacles, dead-shaped obstacles. In such working environments, the robot path planning may easily fall into local optimum or oscillation, resulting that the robot can not find the optimal path. The existing achievements have not solved these problems. For example, the ant colony algorithm for robot path planning just consider the prior information of environment, but don't make posterior analysis for the obstacle's distribution of possible nodes. The algorithm may trap in deadlock after gaining local optimum solution when robot encounters traps in path planning. Therefore, a new ant colony algorithm based on Bayesian decision theory for path planning is presented. In this algorithm, Bayesian model is used in the selection strategy of possible nodes, evaluating the candidate node with posterior probability. This algorithm improves the random search strategy of ant colony algorithm and solves the problem that the standard ant colony algorithm often gets stock into premature stagnation in complex working environment. The results of simulations illustrate that this algorithm has strong search ability, it is very suitable for a complex environment.In an unknown complex environment, the algorithm based on rolling window is commonly used. But oscillation phenomenon is often happened in a trap environment. In order to solve this problem, we propose a new rolling window algorithm based on based on Bayes minimum risk rule for path planning. In this algorithm, we use Bayes minimum risk rule to evaluate the risk of local subgoal's surrounding environment, then select the lowest expected risk of local subgoal as the optimal subgoal. The algorithm solves the problem of rolling window's oscillation. Simulation results show that this algorithm is better than related algorithms in a trap environment, the effect is very satisfactory.