Geolocating minimization of drive test (MDT) measurement reports (MRs) with missing satellite navigation system coordinates
Geolocating Minimization of Drive Test (MDT) measurement reports (MRs) with missing satellite navigation system coordinates is disclosed. In some embodiments, a computing node receives a plurality of complete MRs corresponding to a plurality of user equipments (UEs), wherein each complete MR comprises satellite navigation system coordinates identifying a geographic location of the corresponding UE. The computing node then trains a machine learning (ML) model for estimating UE geographic locations based on the plurality of complete MRs, wherein the ML model maps radio frequency (RF) signatures of complete MRs to corresponding UE geographic locations. In some embodiments, a radio access node obtains the ML model from the computing node, and receives an incomplete MR corresponding to a UE. Upon determining that the second MR lacks satellite navigation system coordinates, the radio access node predicts the geographic location of the UE based on measurements in the incomplete MR and the ML model.
1. A method for geolocating Minimization of Drive Test, MDT, measurement reports, MRs, with missing satellite navigation system coordinates, the method comprising, at a computing node:
receiving a plurality of complete MRs corresponding to a plurality of user equipments, UEs, wherein each complete MR of the plurality of complete MRs comprises satellite navigation system coordinates identifying a geographic location of a corresponding UE of the plurality of UEs; and
training a machine learning, ML, model for estimating UE geographic locations based on the plurality of complete MRs, wherein the ML model maps a plurality of radio frequency, RF, signatures of the plurality of complete MRs to corresponding UE geographic locations.
2. The method of claim 1 , wherein each RF signature of the plurality of RF signatures comprise one or more measurements comprised in a complete MR of the plurality of complete MRs.
3. The method of claim 1 , wherein:
the plurality of complete MRs each further comprises:
a serving cell identifier for a serving cell of the corresponding UE;
a Reference Signal Received Power, RSRP, measurement for the serving cell of the corresponding UE; and
one or more non-serving cell identifiers and corresponding RSRPs for a corresponding one or more non-serving cells measured simultaneously with the serving cell of the corresponding UE; and
training the ML model is further based on the serving cell identifier, the RSRP measurement, and the one or more non-serving cell identifiers and the corresponding RSRPs of each complete MR of the plurality of complete MRS.
4. The method of claim 1 , wherein:
the plurality of complete MRs each further comprises a timing advance, TA, parameter indicating an estimated distance between the corresponding UE and the serving cell of the corresponding UE; and
training the ML model is further based on the estimated distance between the corresponding UE and the serving cell of the corresponding UE indicated by the TA parameter of each complete MR of the plurality of complete MRs.
5. The method of claim 1 , wherein:
the plurality of complete MRs each further comprises a timestamp and a call identifier; and
training the ML model is further based on excluding geolocations based on the timestamp and the call identifier of each complete MR of the plurality of complete MRs.
6. The method of claim 1 , wherein the ML model is based on one or more of a non-linear regression algorithm, a regression tree, and a neural network.
7. The method of claim 1 , wherein each complete MR of the plurality of complete MRs comprises an MDT M 1 report.
8. The method of claim 1 , further comprising, at a radio access node for a Radio Access Network, RAN, of a cellular communications system:
obtaining the ML model from the computing node;
receiving an incomplete MR corresponding to a UE;
determining that the incomplete MR lacks satellite navigation system coordinates identifying a geographic location of the UE; and
responsive to determining that the incomplete MR lacks the satellite navigation system coordinates, predicting the geographic location of the UE based on measurements comprised in the incomplete MR and the ML model.
9. A computing node for geolocating Minimization of Drive Test, MDT, measurement reports, MRs, with missing satellite navigation system coordinates, the computing node comprising:
a network interface; and
processing circuitry adapted to cause the computing node to:
receive a plurality of complete MRs corresponding to a plurality of user equipments, UEs, wherein each complete MR of the plurality of complete MRs comprises satellite navigation system coordinates identifying a geographic location of a corresponding UE of the plurality of UEs; and
train a machine learning, ML, model for estimating UE geographic locations based on the plurality of complete MRs, wherein the ML model maps a plurality of radio frequency, RF, signatures of the plurality of complete MRs to corresponding UE geographic locations.
10. A method performed by a radio access node for a Radio Access Network, RAN, of a cellular communications system to geolocate Minimization of Drive Test, MDT, measurement reports, MRs, with missing satellite navigation system coordinates, the method comprising:
obtaining a machine learning, ML, model for estimating user equipment, UE, geographic locations based on measurements comprised in an MR;
receiving an incomplete MR corresponding to a UE;
determining that the incomplete MR lacks satellite navigation system coordinates identifying a geographic location of the UE; and
responsive to determining that the incomplete MR lacks the satellite navigation system coordinates, predicting the geographic location of the UE based on measurements comprised in the incomplete MR and the ML model.
11. The method of claim 10 , wherein obtaining the ML model for estimating UE geographic locations comprises receiving the ML model from a computing node.
12. The method of claim 10 , wherein obtaining the ML model for estimating UE geographic locations comprises:
receiving a plurality of complete MRs corresponding to a plurality of UEs, wherein each complete MR of the plurality of complete MRs comprises satellite navigation system coordinates identifying a geographic location of a corresponding UE of the plurality of UEs; and
training the ML model for estimating UE geographic locations based on the plurality of complete MRs, wherein the ML model maps a plurality of radio frequency, RF, signatures of the plurality of complete MRs to corresponding UE geographic locations.
13. The method of claim 12 , wherein:
the plurality of complete MRs each further comprises:
a serving cell identifier for a serving cell of the corresponding UE;
a Reference Signal Received Power, RSRP, measurement for the serving cell of the corresponding UE; and
one or more non-serving cell identifiers and corresponding RSRPs for a corresponding one or more non-serving cells measured simultaneously with the serving cell of the corresponding UE; and
training the ML model is further based on the serving cell identifier, the RSRP measurement, and the one or more non-serving cell identifiers and the corresponding RSRPs of each complete MR of the plurality of complete MRs.
14. The method of claim 12 , wherein:
the plurality of complete MRs each further comprises a timing advance, TA, parameter indicating an estimated distance between the corresponding UE and the serving cell of the corresponding UE; and
training the ML model is further based on the estimated distance between the corresponding UE and the serving cell of the corresponding UE indicated by the TA parameter of each complete MR of the plurality of complete MRs.
15. The method of claim 12 , wherein:
the plurality of complete MRs each further comprises a timestamp and a call identifier; and
training the ML model is further based on excluding geolocations based on the timestamp and the call identifier of each complete MR of the plurality of complete MRs.
16. The method of claim 12 , wherein the ML model is based on one or more of a non-linear regression algorithm, a regression tree, and a neural network.
17. The method of claim 12 , wherein each complete MR of the plurality of complete MRs comprises an MDT M1 report.
18. A radio access node for a radio access network, RAN, of a cellular communications system enabled to geolocate Minimization of Drive Test, MDT, measurement reports, MRs, with missing satellite navigation system coordinates, the radio access node comprising:
a network interface; and
processing circuitry associated with the network interface, the processing circuitry adapted to cause the radio access node to:
obtain a machine learning, ML, model for estimating user equipment, UE, geographic locations based on measurements comprised in an MR;
receive an incomplete MR corresponding to a UE;
determine that the incomplete MR lacks satellite navigation system coordinates identifying a geographic location of the UE; and
responsive to determining that the incomplete MR lacks the satellite navigation system coordinates, predict the geographic location of the UE based on measurements comprised in the incomplete MR and the ML model.