Elevator Traffic Forecasting With Bayesian Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Elevator group control systems face inaccuracies in passenger traffic forecasting due to reliance on long-term historical statistics, which fail to adapt to unexpected changes, leading to suboptimal dispatching decisions.
Innovation Solution
A method and apparatus that combine historical traffic information with current cycle data, using a statistical traffic model trained with timestamped origin-destination passenger counts, and employ Bayesian filtering to generate real-time forecasts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Stability of the object's composition
If long-term historical traffic statistics are used for forecasting, then the forecast is stable and consistent, but it fails to adapt to unexpected changes in passenger traffic
Solution Approach 1:
The patent combines long-term historical traffic statistics with short-term observed traffic data to create a hybrid forecasting approach. The controller integrates both data sources, using the stable historical patterns as a baseline while incorporating real-time deviations to adapt to unexpected changes, thus resolving the contradiction between forecast stability and adaptability
Solution Approach 2:
The forecasting system transitions from a static historical model to a dynamic adaptive model. The controller continuously updates the forecast by comparing observed traffic with predicted traffic and adjusting subsequent forecasts based on the deviations, enabling the system to adapt dynamically to changing traffic conditions while maintaining the stabilizing influence of historical data
2Device complexity
If only historical traffic statistics are used, then the forecasting system is simple to implement, but it results in incorrect forecasts when unexpected changes occur
Solution Approach 1:
The patent implements a feedback mechanism where the controller continuously compares observed traffic with predicted traffic, calculates deviations, and uses this feedback to adjust subsequent forecasts. This feedback loop improves forecast accuracy in response to unexpected changes without requiring a complete redesign of the forecasting system, maintaining relative simplicity while enhancing precision
3Adaptability or versatility
If the statistical traffic model is retrained with current cycle data, then the forecast adapts to short-term variations, but the computational requirements and processing time increase
Solution Approach 1:
Instead of completely retraining the statistical traffic model with all current cycle data, the patent applies partial updates by incorporating only the necessary adjustments based on observed deviations. This partial action approach enables the system to adapt to short-term variations while minimizing the time and computational resources required for model updates
Data Source
Figure 1~2
Figure 3~4
AI summary
According to an aspect, there is provided a method for forecasting elevator passenger traffic of an elevator group. The method comprises training (100) a statistical traffic model describing a traffic profile for a specific cycle with historical timestamped origin-destination passenger counts, obtaining (102) timestamped origin-destination passenger counts for a current cycle, generating (104) an elevator passenger traffic forecast based on the trained statistical traffic model and the timestamped origin-destination passenger counts for the current cycle, and outputting (106) the elevator passenger traffic forecast for use by an elevator group control.