Research On Key Issues Of Mobile Robot Navigation In Small Area

With the development of sensor technology, information processing, electronic engineering, computer science, artificial intelligence and other related technology, the performance of robots have been improved continuously. The application has been expanded. At present, robots are not only applied in the industry, agriculture, daily health care, service, but also used to some special fields, such as anti-terrorist, defense and space exploration, military combat and so on.There are a lot of navigation technologies for mobile robot, such as visual navigation, satellite navigation, inertial navigation. They each have advantages and disadvantages, and cannot completely suitable for the small area-navigation in complicated environment, Visual navigation is unstable under condition of random vibration, and cannot be computed real-time. Satellite navigation cannot be used in special environment, such as caves and basement. As the error increasing with time, inertial navigation is not suitable for a long time usage.According to the characteristics of small area navigation in complicated environment, this paper proposes a mobile robot attitude computing scheme based on complementary refactoring of multi-sensor information. Firstly, we design Kalman filters for pitch and roll by using acceleration sensor information to make sure errors is acceptable. Secondly, the filtered pitch, roll and yaw from optical fiber gyroscope are used to reconstruct observations of quaternions. Finally using complementary parameter fuses multi-sensor information and calculates heading and attitude angles. The error of strap-down inertial navigation system is controlled within the3°in quite a time and met the requirement of small area navigation.Based on the characteristics of road of small area in complicated environment, this paper makes use of support vector machine (SVM) to classify the conditions of trajectory, and uses robot motion state to describe operation process. When the robot treks, the position of robot can be achieved quickly according to the robot motion state and road condition of trajectory. Otherwise, this paper proposes a mobile robot motion estimation strategy based on feature point classification. Feature points are divided into far points and near points with the distances between feature points and mobile robot. The far points are sensitive to the rotational movement of robot and used to calculate rotational matrix; the near points are sensitive to translational motion and used to calculate the translational matrix. By using classified feature points, the proposed method can effectively reduce computing time, meanwhile ensure accuracy of motion estimation.Experiments are conducted in complicated outdoor environment. The results show that the methods this paper introduced have good practicability, thus providing an effective solution for mobile robot navigation within small area in complicated environment.