| CPC G01S 5/0264 (2020.05) [G01S 5/02526 (2020.05); H04W 4/021 (2013.01); H04W 4/025 (2013.01)] | 11 Claims |
|
1. A ranging-type positioning method based on crowdsourced calibration, comprising:
arranging a plurality of base stations in a target field;
configuring a mobile device to move in the target field for obtaining direction information and step length information during a movement of the mobile device, and configuring the mobile device to communicate with the plurality of base stations for collecting a plurality of measurement data records related to the plurality of base stations during the movement;
configuring a computing device to:
calculate, according to the plurality of measurement data records, a plurality of measurement distances and a plurality of measurement parameters related to the plurality of measurement distances;
obtain map information of the target field, wherein the map information includes at least one reference coordinate of at least one reference base station among the plurality of base stations;
execute a particle filter for obtaining a forward path, a backward path and a replay path according to the map information, the direction information and the step length information;
combine the backward path and the replay path for generating a mixed path;
perform an optimization process on a ranging model, so as to minimize a distance loss and a geometric loss with respect to one of a plurality of estimated distances generated by the ranging model, wherein the ranging model includes:
an offset compensation module configured to process the plurality of measurement distances through a first activation function and generate a plurality of offset compensation distances, wherein the plurality of offset compensation distances are respectively used to eliminate a plurality of offset errors of the plurality of measurement distances; and
a non-line of sight (NLOS) estimation module configured to, by using a neural network, take the plurality of offset compensation distances and the plurality of measurement parameters as inputs and estimate a plurality of corrected distances, wherein the plurality of corrected distances are used to eliminate a plurality of position-dependent errors caused by NLOS in the plurality of measurement distances;
wherein the distance loss is a sum of differences respectively between the plurality of measurement distances related to the at least one reference base station and a plurality of true distances, and the plurality of true distances are distances between coordinates on the mixed path and the at least one reference coordinate in response to the plurality of measurement distances being obtained;
wherein the geometric loss is a sum of differences respectively between the plurality of measurement distances and a plurality of due distances, and the plurality of due distances are distances respectively between the coordinates on the mixed path and a plurality of to-be-optimized estimated coordinates of the plurality of base stations in response to the plurality of measurement distances being obtained; and
generate, in response to the ranging model being optimized by the optimization process, the plurality of estimated distances, calculate a plurality of estimated coordinates of the base stations according to the plurality of estimated distances, and take the optimized ranging model as a positioning model;
configuring a to-be-positioned mobile device to move in the target field, to obtain positioning direction information and positioning step length information during a movement of the to-be-positioned mobile device, and to communicate with the plurality of base stations for collecting a plurality of positioning measurement data records related to the plurality of base stations during the movement; and
configuring the computing device to:
calculate, according to the plurality of positioning measurement data records, a plurality of positioning measurement distances and a plurality of positioning measurement parameters related to the plurality of positioning measurement distances;
input the plurality of positioning measurement distances and the plurality of positioning measurement parameters into the positioning model, so as to generate a plurality of positioning estimated distances; and
execute, according to the map information, the positioning direction information, the positioning step length information, the estimated coordinates and the positioning estimated distances, the particle filter to obtain a position of the to-be-positioned mobile device in the target field.
|