Federated Learning Scheduling by Route and Base Station Coverage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Federated learning for mobile communication devices, such as connected and autonomous vehicles, faces challenges due to limited communication bandwidth, interference in wireless transmissions, and the need for low latency and high throughput connections.
Innovation Solution
A computer-implemented method and system for scheduling federated learning tasks that involve acquiring data on the routes of mobile communication devices, determining subsets of devices sharing routes, estimating coverage times by base stations, and generating schedules based on priority levels, processing resources, and connectivity metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a centralised server is used for federated learning aggregation, then model training can be performed, but communication delays and network dependency increase
Solution Approach 1:
The patent segments the centralized aggregation function into distributed base station nodes. Instead of all mobile devices communicating with a single central server, the system divides the aggregation task across multiple base stations that serve different geographic areas, reducing communication distance and delay for each device while maintaining the federated learning aggregation capability.
Solution Approach 2:
Base stations act as intermediaries between mobile devices and the central server. Devices communicate model parameters to their serving base station, which then forwards aggregated updates to the central server or performs local aggregation, reducing direct communication delays between devices and the central server while ensuring model training availability.
2Reliability
If more mobile communication devices are scheduled for federated learning updates, then model convergence improves, but communication bandwidth consumption increases
Solution Approach 1:
The patent implements local aggregation at base stations serving specific geographic areas. Each base station aggregates model parameters from devices in its coverage area, processing and compressing data locally before forwarding to the central server. This reduces the total communication bandwidth required compared to all devices transmitting directly to the central server, while still achieving model convergence through distributed aggregation.
Solution Approach 2:
The system schedules devices for federated learning updates in rounds, where only a subset of devices participates in each round rather than all devices simultaneously. This partial action approach allows the model to converge over multiple rounds while controlling bandwidth consumption in each individual round, balancing convergence requirements with available communication resources.
3Productivity
If federated learning tasks are scheduled without considering device routes and base station coverage, then scheduling simplicity is maintained, but transmission reliability and coverage are reduced
Solution Approach 1:
The patent performs preliminary determination of device routes and base station coverage areas before scheduling federated learning tasks. The system identifies which base stations will cover which devices during future time periods, allowing it to pre-assign aggregation responsibilities to appropriate base stations. This preliminary action ensures transmission reliability by guaranteeing that devices will be within coverage of their assigned base station when updates need to be transmitted, while maintaining scheduling efficiency through pre-computed assignments.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There is provided a method comprising: acquiring (110) data associated with the routes of mobile communication devices; determining (120) a subset of mobile communication devices which share a same route for a given amount of time; determining (130) base stations located along the shared route; estimating (140) points of time at which the subset of mobile communication devices are in coverage areas of respective base stations; determining (150) an amount of required processing resources at the base stations and/or at the subset of mobile communication devices; and generating (160) a schedule for a plurality of federated learning tasks to be performed, based on priority levels associated with the federated learning tasks, estimated points of time at which the subset of mobile communication devices are in coverage areas of the base stations, and the amount of required processing resources at the base stations and/or at the subset of mobile communication devices.