Model-assisted deep reinforcement learning based scheduling in wireless networks
In one embodiment, an apparatus of an access point (AP) node of a network includes an interconnect interface to connect the apparatus to one or more components of the AP node and a processor to: access scheduling requests from a plurality of devices, select a subset of the devices for scheduling of resource blocks in a time slot, and schedule wireless resource blocks in the time slot for the subset of devices using a neural network (NN) trained via deep reinforcement learning (DRL).
1. A method comprising:
accessing scheduling requests from a plurality of devices; and
scheduling wireless resource blocks in a plurality of time slots using a neural network (NN) trained via deep reinforcement learning (DRL) to schedule wireless resource blocks for a number K devices, the scheduling for each time slot comprising:
selecting a subset of K devices from the plurality devices; and
scheduling wireless resource blocks in the time slot for the subset of K devices using the DRL-trained NN.
2. The method of claim 1 , wherein the subset of K devices is selected for each time slot using a random round robin selection.
3. The method of claim 1 , wherein the subset of K devices is selected for each time slot based on a sorting of the plurality of devices according to a measure of past throughput for each respective device of the plurality of devices.
4. The method of claim 3 , wherein the measure of past throughput is a ratio of a channel rate for the device and an exponential weighted average rate previously received by the device.
5. The method of claim 1 , wherein the subset of K devices is selected for each time slot based on a sorting of the plurality of devices according to a measure of an amount of data queued for each respective device of the plurality of devices.
6. The method of claim 5 , wherein the measure of the amount of data queued is one of an instantaneous queue length for the device and a time average of queue length for the device.
7. An apparatus of an access point (AP) node of a network, the apparatus including an interconnect interface to connect the apparatus to one or more components of the AP node, and a processor to:
access scheduling requests from a plurality of devices; and
schedule wireless resource blocks in a plurality of time slots using a neural network (NN) trained via deep reinforcement learning (DRL) to schedule wireless resource blocks for a number K devices, wherein the processor is to schedule wireless resource blocks for each time slot by:
selecting a subset of K devices from the plurality devices; and
scheduling wireless resource blocks in the time slot for the subset of K devices using the DRL-trained NN.
8. The apparatus of claim 7 , wherein the processor is to select the subset of K devices for each time slot using a random round robin selection.
9. The apparatus of claim 7 , wherein the processor is to select the subset of K devices for each time slot based on a sorting of the plurality of devices according to a measure of past throughput for each respective device of the plurality of devices.
10. The apparatus of claim 9 , wherein the measure of past throughput is a ratio of a channel rate for the device and an exponential weighted average rate previously received by the device.
11. The apparatus of claim 7 , wherein the processor is to select the subset of K devices for each time slot based on a sorting of the plurality of devices according to a measure of an amount of data queued for each respective device of the plurality of devices.
12. The apparatus of claim 11 , wherein the measure of the amount of data queued is one of an instantaneous queue length for the device and a time average of queue length for the device.
13. The apparatus of claim 7 , wherein the processor is further, for each successive time slot, to:
select a new subset of K devices for scheduling of resource blocks in the time slot; and
schedule wireless resource blocks in the time slot for the new subset of devices using the DRL-trained NN.
14. The apparatus of claim 7 , wherein the processor is further to:
for a number of additional time slots, schedule wireless resource blocks in the additional time slots for the subset of devices using the DRL-trained NN; and
after the number of additional time slots:
select a new subset of K devices for scheduling of resource blocks in a next set of time slots; and
schedule wireless resource blocks in the next set of time slots for the new subset of devices using the DRL-trained NN.
15. The apparatus of claim 7 , wherein inputs to the DRL-trained NN for scheduling wireless resource blocks include on one or more of queue lengths for each of the subset of devices, channel rates for each of the subset of devices, past throughputs for each of the subset of devices, and counters to track last arrivals for each of the subset of devices.
16. The apparatus of claim 7 , wherein the DRL-trained NN is trained to schedule the wireless resource blocks such that an average throughput of the subset of devices is optimized over time.
17. The apparatus of claim 7 , further comprising a wireless transceiver to couple the AP node to an IEEE 802.11-based network.
18. The apparatus of claim 7 , further comprising a wireless transceiver to couple the AP node to a cellular network.
19. One or more non-transitory computer-readable media comprising instructions that, when executed by one or more processors of an access point (AP) node of a network, cause the one or more processors to:
access scheduling requests from a plurality of devices; and
schedule wireless resource blocks in a plurality of time slots using a neural network (NN) trained via deep reinforcement learning (DRL) to schedule wireless resource blocks for a number K devices, the scheduling for each time slot comprising:
selecting a subset of K devices from the plurality devices; and
scheduling wireless resource blocks in the time slot for the subset of K devices using the DRL-trained NN.
20. The computer-readable media of claim 19 , wherein the instructions are to cause the one or more processors to select the subset of K devices for each time slot using a random round robin selection.
21. The computer-readable media of claim 19 , wherein the instructions are to cause the one or more processors to select the subset of K devices for each time slot based on a sorting of the plurality of devices according to a measure of past throughput for each respective device of the plurality of devices.
22. The computer-readable media of claim 21 , wherein the measure of past throughput is a ratio of a channel rate for the device and an exponential weighted average rate previously received by the device.
23. The computer-readable media of claim 19 , wherein the instructions are to cause the one or more processors to select the subset of K devices for each time slot based on a sorting of the plurality of devices according to a measure of an amount of data queued for each respective device of the plurality of devices.
24. The computer-readable media of claim 23 , wherein the measure of the amount of data queued is one of an instantaneous queue length for the device and a time average of queue length for the device.
25. The computer-readable media of claim 19 , wherein the instructions are to further cause the one or more processors, for each successive time slot, to:
select a new subset of K devices for scheduling of resource blocks in the time slot; and
schedule wireless resource blocks in the time slot for the new subset of devices using the DRL-trained NN.