| Simultaneous localization and mapping(SLAM) is the key step to solve the problem that mobile robots truly realize autonomous in an unknown environment. SLAM consists of two parts: 1) State estimation; 2) Data Association. The accuracy of data association is deciding the accuracy of robot localization and map building. The false of data association can lead to errors forecast, and even result in the divergence of the locating and mapping. And correct data is a prerequisite for correct state estimation.In research of SLAM problem, the nearest neighbor association algorithm is the most classic method and is widely used practice application. The core of the nearest neighbor association algorithm is utilizing mahalanobis distance as a condition of compatibility test, so the algorithm is based on simple principle, and is easy to implement However, there are three uncertainties in procedure of calculation SLAM which lead to correct Association rate reduce:1) the associated assumption errors are caused by the equation of robot motion and sensor measurement 2) The measuring error is caused by intensive environment features. 3) Position errors is stacked. Faced with above mentioned problems, the right data Association can not obtained only depended on mahalanobis distance In this thesis, we improved the nearest neighbor algorithm to solve its limitation, and the main study are as follows:1. According to the principle of data associated, the procedure of data association is computing the associated possibility between the features observed by sensor and the features in map. Thus, we introduce conditional probability to calculate the probability between sensor value associated with the map data and the existing features. Then, we analysis of the prediction and observation vector covariance influence mechanism on data association algorithm, proposed the prediction vector and the observation vector covariance auxiliary calculating the mahalanobis distance to improve the nearest neighbor data association algorithm, and through the simulation verify the effectiveness of the algorithm.2. In order to solve the problem that the correct rate is lower in uncertain and complicated environment and prevent false signs, we introduce fuzzy self-adapting algorithm to improve the correct association rate and the accuracy of position. There are errors between message observed by sensor and state estimation of features, so it will improve the result of data association that rule of fuzzy logistic is applied in the data association between the value of observed features and the estimated value. What’s more, we track the least distance between two signs according to observed data in order to prevent false signs. Then, one sign of the least distance is regarded as the new fictitious sign in system, and the other one is utilize to built data association with the fictitious sign.3. Using the two groups of real data which come from the Victoria park and German school laboratory, under the environment of indoor and outdoor experiment respectively to do the experiments. To verify the improved data correlation algorithm is correctnessly and validity which is applied to the SLAM The simulation results show that the improved data correlation algorithm is more suitable for the complicated environment of outdoor conditions. The relevance data accuracy is higher, the anti-interference ability is stronger, the advantages of algorithm is more prominent. |
PDF Full Text Request