Zero-shot deep learning for multi-target RF positioning
Aspects presented herein may enable a passive positioning system, which may be a network entity or node, to be trained to identify multiple moving objects based on using training data for a single object. In one aspect, a network entity receives first RF channel data recorded by a set of devices for a coverage area during a first time period. The network entity trains an ML model based on the set of devices and the first RF channel data. The network entity receives second RF channel data recorded by the set of devices at a second time instance that is outside of the first time period. The network entity computes a number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model.
1 . An apparatus for wireless communication at a network entity, comprising:
at least one memory; and
at least one processor coupled to the at least one memory, and the at least one processor is configured to:
receive first radio frequency (RF) channel data recorded by at least one receiving device in a set of devices for a coverage area during a first time period, wherein the first RF channel data is based on measurement of a first set of RF signals from at least one transmitting device in the set of devices;
train a machine learning (ML) model based on the set of devices and the first RF channel data;
receive second RF channel data recorded by the at least one receiving device in the set of devices at a second time instance that is outside of the first time period, wherein the second RF channel data is based on measurement of a second set of RF signals from the at least one transmitting device in the set of devices; and
compute a number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model, wherein the moving objects exclude the set of devices.
2 . The apparatus of claim 1 , wherein to train the ML model based on the set of devices and the first RF channel data, the at least one processor is configured to:
train the ML model based on superposition of a set of samples randomly selected from the first RF channel data.
3 . The apparatus of claim 2 , wherein the superposition corresponds to a weighted sum of the set of samples, a channel power of each sample in the set of samples being used as a corresponding weight.
4 . The apparatus of claim 1 , wherein the first RF channel data corresponds to a first set of channel state information (CSI) samples and the second RF channel data corresponds to a second set of CSI samples.
5 . The apparatus of claim 1 , wherein the at least one processor is further configured to:
configure the at least one transmitting device to transmit the first set of RF signals; and
configure the at least one receiving device to record the first RF channel data based on the first set of RF signals while one object is moving in the coverage area during the first time period.
6 . The apparatus of claim 5 , wherein the one object corresponds to a single user.
7 . The apparatus of claim 1 , wherein the at least one processor is further configured to:
configure the at least one transmitting device to transmit the first set of RF signals;
configure the at least one receiving device to record the first RF channel data based on the first set of RF signals and a set of locations of an object while the object is moving across the coverage area during the first time period; and
associate the first RF channel data with the set of locations of the object.
8 . The apparatus of claim 7 , wherein the at least one processor is further configured to:
train the ML model further based on the association of the first RF channel data with the set of locations of the object; and
estimate a location for each moving object in the number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model.
9 . The apparatus of claim 7 , wherein the coverage area includes a plurality of rooms, and to configure the at least one receiving device to record the first RF channel data and the set of locations of the object while the object is moving across the coverage area during the first time period, the at least one processor is configured to:
configure the at least one receiving device to record the first RF channel data and the set of locations of the object while the object is moving across the plurality of rooms during the first time period.
10 . The apparatus of claim 9 , wherein each room in the plurality of rooms is associated with a corresponding room identification (ID) or a corresponding room index, and to associate the first RF channel data with the set of locations of the object, the at least one processor is configured to:
associate each subset of the first RF channel data with the corresponding room ID or the corresponding room index based on a set of inputs received from the object.
11 . The apparatus of claim 1 , wherein to train the ML model based on the set of devices and the first RF channel data, the at least one processor is configured to:
decompose a superposition of the first RF channel data into a first plurality of channel state information (CSI) samples, wherein each CSI sample in the first plurality of CSI samples is associated with a pattern of a single object in the number of moving objects; and
store the first plurality of CSI samples and their corresponding patterns for the single object in a database.
12 . The apparatus of claim 11 , wherein to compute the number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model, the at least one processor is configured to:
decompose a superposition of the second RF channel data into a second plurality of CSI samples; and
identify the number of moving objects based on comparing the second plurality of CSI samples with the first plurality of CSI samples in the database.
13 . The apparatus of claim 12 , wherein to decompose the superposition of the first RF channel data and the superposition of the second RF channel data, the at least one processor is configured to decompose the superposition of the first RF channel data and the superposition of the second RF channel data using a source separation mechanism.
14 . The apparatus of claim 1 , wherein the first RF channel data and the second RF channel data are associated with a Wi-Fi channel, a Bluetooth channel, an ultra-wideband (UWB) channel, or a network channel.
15 . A method of wireless communication at a network entity, comprising:
receiving first radio frequency (RF) channel data recorded by at least one receiving device in a set of devices for a coverage area during a first time period, wherein the first RF channel data is based on measurement of a first set of RF signals from at least one transmitting device in the set of devices;
training a machine learning (ML) model based on the set of devices and the first RF channel data;
receiving second RF channel data recorded by the at least one receiving device in the set of devices at a second time instance that is outside of the first time period, wherein the second RF channel data is based on measurement of a second set of RF signals from the at least one transmitting device in the set of devices; and
computing a number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model, wherein the moving objects exclude the set of devices.
16 . The method of claim 15 , wherein training the ML model based on the set of devices and the first RF channel data comprises:
training the ML model based on superposition of a set of samples randomly selected from the first RF channel data.
17 . The method of claim 16 , wherein the superposition corresponds to a weighted sum of the set of samples, a channel power of each sample in the set of samples being used as a corresponding weight.
18 . The method of claim 15 , wherein the first RF channel data corresponds to a first set of channel state information (CSI) samples and the second RF channel data corresponds to a second set of CSI samples.
19 . The method of claim 15 , further comprising:
configuring the at least one transmitting device to transmit the first set of RF signals; and
configuring the at least one receiving device to record the first RF channel data based on the first set of RF signals while one object is moving in the coverage area during the first time period.
20 . The method of claim 19 , wherein the one object corresponds to a single user.
21 . The method of claim 15 , further comprising:
configuring the at least one transmitting device to transmit the first set of RF signals;
configuring the at least one receiving device to record the first RF channel data based on the first set of RF signals and a set of locations of an object while the object is moving across the coverage area during the first time period; and
associating the first RF channel data with the set of locations of the object.
22 . The method of claim 21 , further comprising:
training the ML model further based on the association of the first RF channel data with the set of locations of the object; and
estimating a location for each moving object in the number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model.
23 . The method of claim 21 , wherein the coverage area includes a plurality of rooms, and wherein configuring the at least one receiving device to record the first RF channel data and the set of locations of the object while the object is moving across the coverage area during the first time period comprises:
configuring the at least one receiving device to record the first RF channel data and the set of locations of the object while the object is moving across the plurality of rooms during the first time period.
24 . The method of claim 23 , wherein each room in the plurality of rooms is associated with a corresponding room identification (ID) or a corresponding room index, and wherein associating the first RF channel data with the set of locations of the object comprises:
associating each subset of the first RF channel data with the corresponding room ID or the corresponding room index based on a set of inputs received from the object.
25 . The method of claim 15 , wherein training the ML model based on the set of devices and the first RF channel data comprises:
decomposing a superposition of the first RF channel data into a first plurality of channel state information (CSI) samples, wherein each CSI sample in the first plurality of CSI samples is associated with a pattern of a single object in the number of moving objects; and
storing the first plurality of CSI samples and their corresponding patterns for the single object in a database.
26 . The method of claim 25 , wherein computing the number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model comprises:
decomposing a superposition of the second RF channel data into a second plurality of CSI samples based on a source separation mechanism; and
identifying the number of moving objects based on comparing the second plurality of CSI samples with the first plurality of CSI samples in the database.
27 . An apparatus for wireless communication at a network entity, comprising:
means for receiving first radio frequency (RF) channel data recorded by at least one receiving device in a set of devices for a coverage area during a first time period, wherein the first RF channel data is based on measurement of a first set of RF signals from at least one transmitting device in the set of devices;
means for training a machine learning (ML) model based on the set of devices and the first RF channel data;
means for receiving second RF channel data recorded by the at least one receiving device in the set of devices at a second time instance that is outside of the first time period, wherein the second RF channel data is based on measurement of a second set of RF signals from the at least one transmitting device in the set of devices; and
means for computing a number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model, wherein the moving objects exclude the set of devices.
28 . A non-transitory computer-readable medium storing computer executable code at a network entity, the code when executed by a processor causes the processor to:
receive first radio frequency (RF) channel data recorded by at least one receiving device in a set of devices for a coverage area during a first time period, wherein the first RF channel data is based on measurement of a first set of RF signals from at least one transmitting device in the set of devices;
train a machine learning (ML) model based on the set of devices and the first RF channel data;
receive second RF channel data recorded by the at least one receiving device in the set of devices at a second time instance that is outside of the first time period, wherein the second RF channel data is based on measurement of a second set of RF signals from the at least one transmitting device in the set of devices; and
compute a number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model, wherein the moving objects exclude the set of devices.