Predictive Time Allocation Scheduling for LLN Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deterministic networking in low power and lossy networks (LLNs) faces challenges such as lossy links, low bandwidth, and scalability issues, requiring efficient time slot allocation to maintain precise packet delivery and avoid congestion, especially with changing environmental conditions and traffic patterns.
Innovation Solution
A machine learning-based predictive time allocation scheduling system that adjusts time slot assignments in channel hopping schedules based on predicted traffic changes and seasonality, using reports from network nodes to proactively allocate or reallocate time slots, allowing for efficient handling of deterministic and non-deterministic traffic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional time slot allocation methods are used in deterministic networking, then packet delivery precision can be maintained, but network scalability and adaptability to changing traffic patterns deteriorate
Solution Approach 1:
The patent implements dynamic time slot allocation where the network node continuously monitors traffic patterns and adjusts time slot assignments in real-time. The scheduling is no longer static but adapts to changing traffic conditions, allowing the system to maintain precision while becoming versatile across different traffic scenarios
Solution Approach 2:
The patent employs feedback mechanisms where the network node receives information about actual traffic patterns and uses this feedback to adjust future time slot allocations. This closed-loop approach enables the system to learn from past performance and continuously optimize packet delivery precision while adapting to new traffic patterns
2Device complexity
If fixed time slot assignments are used, then control plane overhead is reduced, but network performance under varying traffic demands deteriorates
Solution Approach 1:
The patent uses machine learning models to predict future traffic patterns and proactively adjusts time slot allocations before traffic demands actually change. This preliminary action allows the system to prepare optimal schedules in advance, maintaining high performance while avoiding the need for complex real-time control plane interventions
3Adaptability or versatility
If machine learning-based predictive scheduling is implemented, then adaptability to traffic patterns improves, but device complexity and processing requirements worsen
Solution Approach 1:
The patent implements lightweight machine learning models that can be executed on constrained network nodes. These simplified models provide the necessary adaptability while consuming minimal processing resources, effectively serving as 'cheap' computational solutions that enable complex behavior without requiring high-end hardware
Data Source
AI summary
In one embodiment, a device in a network receives data regarding traffic volumes of deterministic and non-deterministic traffic along a first path in the network. The device predicts, using the received data, an increase in the traffic volume of the non-deterministic traffic along the first path in the network. The device identifies a period of time associated with the predicted increase in the traffic volume of the non-deterministic traffic along the first path. The device causes the deterministic traffic to be sent along a second path in the network during the identified period of time, to allow the first path to accommodate the predicted increase in the traffic volume of the non-deterministic traffic along the first path.