Intelligent Network Routing via Reward-Based Policy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for routing items through networks do not always select the most efficient route in real-time, leading to delays due to traffic and other factors, necessitating a method that utilizes historical data to intelligently route items based on delay analysis.
Innovation Solution
A routing method that evaluates each segment of a trip by assigning small negative rewards for delays and large positive rewards for completion, using these rewards to calculate the value of each trip and define a policy for selecting the best route with the highest net reward, thereby maximizing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional routing methods are used, then the routing system is simple to implement, but the route selection efficiency deteriorates due to inability to adapt to real-time traffic conditions
Solution Approach 1:
The patent implements feedback by using historical trip data and delay information to continuously improve routing decisions. The system collects data from completed trips, analyzes delays at different segments, and uses this feedback to refine future route selections through the value iteration algorithm, creating a closed-loop system that learns from past performance
Solution Approach 2:
The patent applies preliminary action by pre-calculating route values and building a policy database before actual routing decisions are needed. The value iteration algorithm pre-computes the expected delays and optimal routes based on historical data, so when a routing decision is required, the system can quickly query the pre-computed policy rather than calculating from scratch
2Loss of time
If real-time adaptive routing is implemented, then the delivery time is reduced through optimized path selection, but the computational complexity increases due to historical data processing
Solution Approach 1:
The patent changes parameters by transforming raw historical delay data into normalized route values through the value iteration algorithm. Instead of processing raw traffic data directly, the system converts delay measurements into expected delay values and route scores, making the computational problem more tractable and enabling efficient real-time queries
Solution Approach 2:
The patent uses copying by creating a simplified model of the network topology and routing options that mirrors the actual network structure. This computational model allows the value iteration algorithm to evaluate routes efficiently without requiring complex real-time simulations, separating the computational burden from actual routing operations
3Measurement precision
If historical data is utilized for route selection, then the routing accuracy is improved through delay pattern recognition, but the data processing requirements increase
Solution Approach 1:
The patent extracts only the essential delay information from historical trip data that is relevant for routing decisions. Instead of processing all raw data including irrelevant details, the system extracts delay measurements by route segment and destination, filtering out unnecessary information and focusing computational resources on the most predictive features
Data Source
AI summary
A method for selecting routes for items being sent from a given beginning node to a given destination node within a network. Each segment of a trip is evaluated and assigned a negative reward based on the deviation from the ideal length of time to complete that segment. Upon arrival of the item at the destination node, the route is assigned a large positive reward. The system is configured to attempt to maximize rewards at all times by selecting the routes with the lowest likelihood of delay. As more and more trips are completed, the control system may define a policy that selects routes based on their probability of yielding the highest reward.