Remaining-Time Forecasting With Confidence-Based Smoothing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing residual value forecasting methods for systems like traffic lights are unreliable due to insufficient consideration of factors such as statistical uncertainties and unforeseen events, leading to unpredictable fluctuations in the countdown of remaining time until a switching point.
Innovation Solution
A method that adjusts the current remaining time value based on a series of previous values and their associated confidences, using linear extrapolation and neural network outputs to smooth the forecast, ensuring reliable and consistent countdown.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional neural model output is used for residual value forecast, then the forecast can be generated quickly, but the reliability is low due to insufficient consideration of statistical uncertainties and unforeseen events
Solution Approach 1:
The forecast system is segmented into multiple independent components: individual remaining time forecasts from neural networks, separate confidence values, and distinct adjustment mechanisms. This segmentation allows each component to be optimized independently while collectively improving reliability without sacrificing generation speed.
Solution Approach 2:
The system implements feedback by using confidence values as a mechanism to adjust remaining time forecasts. The adjustment process feeds back into the forecast output, creating a self-correcting system that improves reliability by accounting for uncertainties and deviations from previous forecasts.
2Reliability
If confidence-based adjustment is applied to smooth fluctuations, then user confidence improves, but the system complexity increases due to multiple processing steps
Solution Approach 1:
The system changes parameters by introducing confidence values as an additional dimension for forecast evaluation. By adjusting the remaining time based on confidence levels, the system smooths fluctuations while maintaining a relatively simple processing architecture that doesn't significantly increase overall system complexity.
3Stability of the object's composition
If multiple previous remaining time values are considered for adjustment, then forecast stability improves, but the processing time increases
Solution Approach 1:
The system applies partial action by considering only the necessary number of previous forecasts and their confidence values for adjustment. It processes enough historical data to achieve stability but avoids excessive processing by implementing efficient adjustment algorithms that balance stability improvement with processing time constraints.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for determining an adjusted current remaining time value for a plant, comprising the steps of: a. providing a plurality of previous remaining time values, each with at least one associated confidence level, within a specific period (S1); b. selecting at least one most recent previous remaining time value from the plurality of previous remaining time values, which immediately precedes the current remaining time value within the specific period; c. receiving a current remaining time value with at least one current confidence level (S2); and d. determining the adjusted current remaining time value based on the plurality of previous remaining time values with their respective associated confidence levels, depending on the current confidence level (S3). The invention further relates to a corresponding computer program and a determination unit.