Methods, apparatus, and systems for artificial intelligence (AI)-enabled filters in wireless systems
Methods, apparatus and systems are disclosed. One method may include a wireless transmit/receive unit (WTRU) receiving a transmission including a data unit (DU) on a first set of resources. The WTRU may select an artificial intelligence (AI) filter based on the first set of resources and input the DU or a part of the DU to the selected AI filter. The WTRU may perform AI filtering on the inputted DU or part thereof to output any of: a set of AI-based transmission parameters or an AI-processed DU. The AI-processed DU may include: a first portion of the DU processed by the AI filter and a second portion of the DU processed by a rule-based component, or the DU processed by the AI filter. The WTRU may transmit any of: the AI-processed DU using a set of rule-based transmission parameters, or a rule-based DU using the AI-based transmission parameters.
1 . A method implemented by a wireless transmit/receive unit (WTRU), the method comprising:
determining transmission resources and meta information associated with the transmission resources;
determining a first Artificial Intelligence (AI) filter from one or more AI filters based on the meta information associated with the transmission resources and contextual information;
receiving interface information indicating any of: (1) an entry point of the first AI filter, (2) initial values or parametrization for the first AI filter, and (3) an exit point of the first AI filter;
based on the interface information, applying, as an input to the first AI filter, information regarding: (1) one or more of the determined transmission resources, (2) the meta information associated with the transmission resources, and (3) one or more packet data unit (PDU) headers or a portion of the PDU headers;
obtaining an output of the first AI filter;
obtaining, based on the output of the first AI filter, any of (1) a set of AI-determined transmission parameters or (2) one or more AI-determined processed data units; and
transmitting: (1) at least one of the AI-determined processed data units or (2) a rule-determined processed data unit, using at least one of the AI-determined transmission parameters.
2 . The method of claim 1 , wherein the transmission resources are scheduled transmission resources or configured transmission resources.
3 . The method of claim 1 , wherein the contextual information includes information associated with any of: (1) historical channel conditions, (2) service mix, (3) temporal characteristics of PDUs in a buffer, and/or (4) available processing power at the WTRU.
4 . The method of claim 1 , wherein:
the AI filter includes a memory; and
the method further comprises receiving, from a network entity, a control signal to reset the memory.
5 . The method of claim 1 , wherein:
the AI filter includes a memory to store values associated with weights and biases; and
the method further comprising receiving, from a network entity, a control signal to reset the memory to restore values of the weights and biases to default values.
6 . The method of claim 1 , wherein:
the AI filter uses a plurality of AI nodes; and
the WTRU includes a memory to store (1) an AI filter model of the AI filter, (2) a plurality of weight values associated with the plurality of AI nodes of the AI filter, and (3) a plurality of bias values associated with the plurality of AI nodes of the AI filter.
7 . The method of claim 1 , wherein the input applied to the AI filter includes information regarding any one or more of: (1) one or more statuses of previous transmissions, and (2) one or more historical channel state conditions.
8 . The method of claim 1 , further comprising receiving information to configure the AI filter via any of: a unicast transmission, a broadcast transmission and/or multicast transmission.
9 . The method of claim 1 , further comprising:
receiving information indicating to disable the AI filter; and
disabling the AI filter based on the received information.
10 . The method of claim 9 , further comprising, on condition that the AI filter is disabled, performing a corresponding rule-based processing operation in substitution for the disabled AI filter.
11 . The method of claim 1 , wherein:
the AI filter includes a neural network; and
the method further comprises:
receiving information indicating a set of weights and/or biases to performing AI filtering, and
training the AI filter by setting the weights and/or biases of each neural network node of the neural network based on the received information.
12 . The method of claim 1 , further comprising receiving information to configure the AI filter via a Media Access Control (MAC) Control Element (CE), or downlink control information (DCI).
13 . The method of claim 1 , further comprising:
performing AI-based Logical Channel Prioritization (LCP) using the AI filter; and
generating the AI-processed packet data unit in accordance with the AI-based LCP.
14 . A wireless transmit/receive unit (WTRU), comprising:
circuitry, including any of a processor and a transmit/receive unit, configured to:
determine transmission resources and meta information associated with the transmission resources,
determine a first Artificial Intelligence (AI) filter from one or more AI filters based on the meta information associated with the transmission resources and contextual information,
receive interface information indicating any of: (1) an entry point of the first AI filter, (2) initial values or parametrization for the first AI filter, and (3) an exit point of the first AI filter;
based on the interface information, apply, as an input to the first AI filter, information regarding: (1) one or more of the determined transmission resources, (2) the meta information associated with the transmission resources, and (3) one or more packet data unit (PDU) headers or a portion of the PDU headers,
obtain an output of the first AI filter, and
obtain, based on the output of the first AI filter, any of (1) a set of AI-determined transmission parameters or (2) one or more AI-determined processed data units; and
transmit: (1) at least one of the AI-determined processed data units or (2) a rule-determined processed data unit, using at least one of the AI-determined transmission parameters.
15 . The WTRU of claim 14 , wherein the transmission resources are scheduled transmission resources or configured transmission resources.
16 . The WTRU of claim 14 , wherein the contextual information includes information associated with any of: (1) historical channel conditions, (2) service mix, (3) temporal characteristics of PDUs in a buffer, and/or (4) available processing power at the WTRU.
17 . The WTRU of claim 14 , wherein:
the AI filter includes a memory; and
the transmit/receive unit is configured to receive, from a network entity, a control signal to reset the memory.
18 . The WTRU of claim 14 , wherein:
the AI filter includes memory to store values associated with weights and biases; and
the transmit/receive unit is configured to receive, from a network entity, a control signal to reset the memory to restore values of the weights and biases to default values.
19 . The WTRU of claim 14 , wherein the AI filter uses a plurality AI nodes, wherein the WTRU further comprises a memory to store (1) an AI filter model of the AI filter, (2) a plurality of weight values associated with the plurality of AI nodes of the AI filter, and (3) a plurality of bias values associated with the plurality of AI nodes of the AI filter.
20 . The WTRU of claim 14 , wherein the input applied to the AI filter further includes information regarding any one or more of: (1) one or more statuses of previous transmissions, and (2) one or more historical channel state conditions.
21 . The WTRU of claim 14 , wherein the transmit/receive unit is configured to receive information to configure the AI filter via any of: a unicast transmission, a broadcast transmission and/or a multicast transmission.
22 . The WTRU of claim 14 , wherein:
the transmit/receive unit is configured to receive information indicating to disable the AI filter; and
the processor is configured to disable the AI filter based on the received information.
23 . The WTRU of claim 22 , wherein the processor is configured to perform, on condition that the AI filter is disabled, a corresponding rule-based processing operation in substitution for the disabled AI filter.
24 . The WTRU of claim 14 , wherein:
the AI filter includes a neural network; and
the transmit/receive unit is configured to receive information indicating a set of weights and/or biases to perform AI filtering, and
the processor is configured to train the AI filter by setting the weights and/or biases of each neural network node of the neural network based on the received information.
25 . The WTRU of claim 14 , wherein the transmit/receive unit is configured to receive information to configure the AI filter via a Media Access Control (MAC) Control Element (CE), or downlink control information (DCI).
26 . The WTRU of claim 14 , wherein the processor is configured to:
perform AI-based Logical Channel Prioritization (LCP) using the AI filter; and
generate the AI-processed packet data unit in accordance with the AI-based LCP.
27 . The method of claim 1 , wherein the input to the first AI filter further comprises information regarding any of: link quality and one or more logical channel identities.
28 . The WTRU of claim 14 , wherein the input to the first AI filter further comprises information regarding any of: link quality and one or more logical channel identities.