Research On Robot Simultaneous Location Mapping And Multi - Target Tracking Algorithm

The Simultaneous Localization and Mapping (SLAM) problem asks if it is possible for a mobile robot to be placed at an unknown location in an unknown environment and for the robot to incrementally build a consistent map of this environment while simultaneously determining its location within this map. A Multi-Target Tracking algorithm jointly estimates the number of targets and their states using the collection of measurements reported by the sensor at each time step. Introduction of random finite set to solving the SLAM problem and Multi-Target Tracking problem have attracted great attention in recent years.SLAM and Multi-Target Tracking are closely related as they both estimate states in the way of a Bayesian estimation problem, sharing similar solutions. Improved solutions to SLAM and Multi-Target Tracking are proposed here, and a random finite set and classification tree based multi-target state extraction approach is introduced. Main contributions are as follows:1. Proposing a wavelet transform assisted extended Kalman filter-based approach for SLAM, which attempts to estimate the elements of measurement noise variance through wavelet transform and uses them to calculate the variance of innovation. The proposed approach is able to achieve SLAM in situations where measurement noise variance is incorrect.2. Proposing an improved Cardinalized Probability Hypothesis Density Filter. This algorithm chooses appropriate formula according to the estimated number of targets when updating density function. This approach combines the merits of Probability Hypothesis Density filter and Cardinalized Probability Hypothesis Density PHD filter. It can both follow the change of the number of targets quickly and get accurate results.3. Proposing a random finite set and classification tree based multi-target state extraction approach. This approach calculates existence factor vector according to information from3time steps and separate them into two categories:"existent" and "non-existent". Pick out the targets of "existent" existence factor vectors and more accurate multi-target states are extracted.Simulations compare the performance of these algorithms with others and have proved their effectiveness and advantage.