Universal Path Planning Based on Naive Bayes Classifier and Improved A* Algorithm using CNN

As the basis of robot kinematics, path planning occupies an important position in artificial intelligence and other fields. However, not many researches have focus on the importance of generating waypoints in path planning. In order to accurately find the most efficient moving path, we proposes a path planning method based on Naive Bayes Classifier and CNN to improve A* algorithm, which is a search-based algorithm. It first constructs a cost map of the space where the target object is located, and obtains the starting point, ending point with path contour. Next, we obtain the contours of obstacles to calculate the size, use Naive Bayes to realize the mapping, and update the cost map faster instead of the classifers with network structure that need supervised training. After that, we use CNN to find key waypoints, eliminate redundant waypoints and calculate the robot gait. Finally the path planning of the robot is realized. Experiments were carried out in the simulation environment Gazebo and the real space respectively, and the results showed the effectiveness and feasibility of the method.