ETA Refinement via Locality-Sensitive Hashing and Transformer Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for computing estimated time of arrival (ETA) in vehicle travel are inaccurate due to route uncertainty, human error, and distribution shifts, as they rely on graph-based models that fail to account for real-world driver decisions and varying task types.
Innovation Solution
A machine-learned model is implemented to refine ETAs by using geospatial features transformed with locality-sensitive hashing, a transformer architecture with self-attention, and a bias adjustment layer, treating the routing engine's ETA as a noisy estimate and correcting it with observational data to align with real-world outcomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If graph-based shortest-path algorithms are used to compute ETA, then computational efficiency is improved, but measurement precision deteriorates due to inability to capture real-world conditions
Solution Approach 1:
The patent introduces a machine-learned model as an intermediary between the graph-based routing algorithm and the final ETA output. This model takes the predicted ETA and additional features as input, then refines the prediction by learning from historical observational data. The intermediary model corrects systematic errors in the graph-based approach while maintaining computational efficiency, as it processes results rather than replacing the entire routing system.
Solution Approach 2:
The patent transforms the ETA prediction problem by changing the parameters fed into the prediction system. Instead of only using graph-based features, the system incorporates additional features (traffic conditions, weather, historical data) and applies a machine-learned transformation. The model learns optimal parameter combinations and transformations from data, enabling more accurate predictions without sacrificing the speed of graph-based routing.
2Device complexity
If conventional routing algorithms are used, then device complexity is reduced, but reliability deteriorates due to route uncertainty and human error
Solution Approach 1:
The system implements feedback by using historical observational data to train the machine-learned model. The model learns from past predictions and actual outcomes, continuously improving its ability to correct routing algorithm errors. This feedback mechanism enhances reliability by incorporating real-world performance patterns without requiring complex real-time adjustments or human intervention.
3Measurement precision
If machine-learned models with multiple features are used to refine ETA, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent segments the ETA prediction system into two distinct components: a graph-based routing algorithm for computing baseline paths and a machine-learned model for refinement. This segmentation allows each component to specialize - the routing algorithm handles path finding efficiently while the ML model focuses on accuracy improvement. The modular architecture manages complexity by separating concerns and allowing independent optimization of each segment.
Data Source
AI summary
An estimated time of arrival (ETA) of a vehicle is predicted by receiving a request for the vehicle to conduct a trip that includes a first location. A predicted ETA for the vehicle to travel from a particular location to the first location is computed. The predicted ETA is refined to compute a refined ETA using a machine-learned model that takes as input a plurality of features associated with the trip. The plurality of features including at least geospatial features transformed using a locality-sensitive hashing function. An action is performed based on the refined ETA. The action may include one or more of estimating a pickup time or drop-off time for the trip, matching a driver to the trip, and planning a delivery.