Environmental Modeling And Navigation Of The Robot Based On Multi-Sensors

Robot environmental modeling and obstacle avoidance and navigation are important research in the field of robotics, this area has attracted a lot of attention of many researchers, and most of them have made great success. However, it is worth noting that most of these methods are limited to the theoretical analysis or simulation, they don’t combine with real robots and real environment, so the methods have a certain distance to the practical. Therefore, this paper focuses on the robot in a real environment, including the methods on environmental modeling, the technology of obstacle avoidance and navigation.For the environmental modeling in the actual environment, the robot should use sensors to detect the environment around it, and model the environment by using the information. If the robot uses a single sensor to detect the environment, the information may be incomplete, and it will lead some errors in the model and other problems. So, this paper proposes a method that combining the laser sensor and visual sensor to detect the local unknown environment, the approximate obstacles are determined by the visual sensor in the local area, the laser sensor detects the distance, and get the coordinate of the barriers determine by the combination of the two kinds of sensor at last. Based on this, the venue will be divided into different grids, the areas which have fewer obstacles are described by the large grid, and the area having more obstacles is described by a small grid. This adaptive grid modeling approach can not only reduce the amount of computation, but also raise the accuracy of the environment model. The experiment results show that the global model build by the adaptive grid has greatly improved in the storage space, precision and efficiency compared with the traditional modeling methods. The local environment model which builds by the combination of laser sensor and visual sensor is better in the environmental integrity, modeling accuracy compared with a single sensor modeling result.The majority of existing robot navigation methods only uses the simulation for path planning; the robot gets the navigation results based on the simulation. During the actual navigation, the path planning methods need to establish a precise environment model, this will result in computationally intensive, and the robot sensor’s detection range is limited, it is difficult to get a global map of the environment. Thus, this paper presents an obstacle avoidance navigation method. In the robot actual work environment, in order to establish the local environment model, we use the laser sensor and visual sensor to detect obstacles in the local environment. VFH*local obstacle avoidance algorithm is applied on local model to complete the navigation. By this way, when the robot selects the walking path, it will be no longer confine to a single sensor detection range and the current movement window, it can select a better path from wider range. The experiment results show that the robot avoids obstacles quickly and safely and it also choose a good path to the target during navigation.