The GOODPUT System: A Machine Learning-Driven Optimization Framework for Dynamic...
summary
Optimizes application-layer goodput by jointly choosing AP channel assignments, channel bonding decisions, and station→AP mappings using LSTM demand predictions and real-time spectrum sensing to build MILP instances solved to optimality; prototype runs on 3 APs, 15 hosts and 100 virtual stations, uses an LSTM demand predictor, configures APs via selenium web driver and steers clients with WPACI, increases goodput +17.1% vs load‑balancing, +19.3% vs ML AP selection, +47.9% vs RSSI and cuts spectrum use ~14% (down to 200 MHz after 482 s).
AI-written summary · about
chapters (8)
- 0:00GOODPUT System intro
- 1:23Wireless network challenges
- 2:25Why existing solutions fail
- 3:27Supply-demand model & OFDMA
- 5:33Pipeline: sensing, optimization, control
- 7:41Hardware testbed setup
- 8:45Experimental results & gains
- 10:11Spectrum utilization & interference
related talks
18mMidgress-aware traffic provisioning for content delivery
13mAUTO: Adaptive Congestion Control Based on Multi-Objective Reinforcement Learning...
16mTowards Optimal Rack-scale μs-level CPU Scheduling through In-Network Workload...
17mDOTE: Rethinking (Predictive) WAN Traffic Engineering
21mFrom Reactive to Proactive: How AI is Redefining Resilience in Modern Networks
12mCATO: End-to-End Optimization of ML-Based Traffic Analysis Pipelines
10mThe Plan That Learns
12mATP: In-network Aggregation for Multi-tenant Learning