Indoor robot positioning method based on improved particle filtering

Abstract

An indoor robot positioning method based on improved particle filtering comprises the steps that S1, extracting parameters, wherein the parameters comprise a particle set at the previous moment, mileage information, measurement information, a map, the number of particles at the previous moment, the number of particles at the previous moment before last and an updating trend of the position of theparticles at the previous moment; S2, updating each particle in the particle set at the previous moment, and fusing the updating trend at the previous moment, to obtain a new particle set; S3, judgingthe updated particle set, and judging (please refer to the formula in the specification) is true or not, and if yes, executing an S4; if not, xd=yd=0; s5, executing a step S5; S4, calculating the updating trend of the position of the state particles at the moment; S5, calculating the number of the particles at the current moment; and S6, outputting a result. According to the method, calculation is optimized while particle updating is carried out, the robot positioning repair can be rapidly converged to the correct position, and the indoor positioning requirement of the robot can be met.

Description

technical field [0001] The invention relates to the field of robot positioning, and more particularly, relates to an indoor robot positioning method based on improved particle filtering. Background technique [0002] Positioning plays an important role when the robot performs tasks; real-time mapping, path planning and automatic obstacle avoidance all require accurate geographical coordinates of the robot in the world coordinates, and its probability-based positioning algorithms mainly include the extended Kalman filter algorithm and the unscented Kalman filter algorithm. Mann filter algorithm, histogram filter algorithm and particle filter algorithm, the first two need feature-based landmark assistance, not suitable for global positioning, and the quality of histogram filter depends heavily on the roughness of the grid map, for ordinary low-cost hardware For robots, a finer map means longer computation times. [0003] The particle filter implements the recursive Bayesian f...

AI Technical Summary

Problems solved by technology

[0003]The particle filter implements the recursive Bayesian filter algorithm using the Monte Carlo mechanism; it is not limited by non-Gaussian noise and nonlinear models, and maintains particle collections Xt to approximate the posterior confidence bel(xt); currently, the optimization of particle filter algorithm mainly focuses on reducing the number of particles and particle convergence speed in the filtering process The improvement of the particle’s anti-interference ability and other aspects; Liu et al. implement the residual strategy to only update some key particles, while the rest of the particles remain unchanged, reducing the redundancy in calculation, but only updating some particles reduces the algorithm’s anti-interference ability. Interference ability, Li T, etc. considered the spatial information of particles, and used the grouping strategy to merge particles to maintain the diversity of particles, but the calculation and merging process took up a considerable overhead; some optimization methods use parallel distributed computing, using GPU improves the calculation speed for particle filter, but this improvement requires too much hardware; others adjust the number of particles required for each iteration in particle filter as much as possible by fitting the real posterior; but because most indoor robots Small in size and low in cost, the improvement research mentioned above is difficult to meet the needs of indoor positioning

Images