Last-Segment Navigation Using UGC and Courier Behavior
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing navigation systems face challenges in accurately determining the last segment of a route, particularly in large complexes like apartment buildings or malls, due to data coverage gaps and slow data refresh rates in third-party data, leading to uncertainties in parking and entrance locations.
Innovation Solution
A system utilizing user-generated content (UGC) and machine learning to enhance location data accuracy by allowing users to input recommended parking and entrance locations, which are then integrated with historical trip data to train a machine learning model for improved navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If third-party data is used to obtain last segment information, then location data coverage is improved, but data refresh rate becomes slow and data coverage gaps remain
Solution Approach 1:
The system enables users (couriers and customers) to self-report last segment information such as parking locations, entrance locations, and delivery points directly through the platform. This user-generated content approach eliminates dependence on third-party data providers, allowing the system to continuously refresh location data with real-time user contributions while maintaining comprehensive coverage across all markets.
2Measurement precision
If inference data from past delivery trips is used, then some location information is obtained, but data points are insufficient leading to inaccurate predictions
Solution Approach 1:
The system merges multiple data sources including user-generated content, historical delivery trip data, and third-party data to create a comprehensive location database. By combining these diverse data streams, the system accumulates sufficient data points to accurately determine last segment information even for locations with limited historical deliveries, thereby improving measurement precision.
3Ease of operation
If traditional navigation systems are used for last segment navigation, then basic routing is provided, but accuracy in complex environments like large buildings is poor
Solution Approach 1:
The navigation system is segmented into multiple components: traditional GPS-based routing for the main route, and a specialized last segment navigation module that provides detailed turn-by-turn directions for the final delivery point. This segmentation allows the system to use simplified methods for long-distance navigation while applying specialized, high-precision techniques for the critical last segment in complex environments.
Data Source
AI summary
Example implementations are directed to systems and methods for improving navigation accuracy for a last segment of a delivery route. A client application on a user device is configured to display user interfaces that allow users to provide user-generated content (UGC) to refine last segment data, such as parking locations, building entrances, and drop-off points. The UGC is collected via interactive map-based tools, where users can adjust pins and provide metadata including entry codes and images. The system integrates the UGC with historical trip data and inference data to generate updated last segment data, which is presented to couriers. Conflicts between the UGC and the inference data can be resolved by analyzing courier behavior and prioritizing the data source most frequently followed. A machine learning model can also be retrained using the UGC, inference data, and courier behavior to improve future predictions of the last segment data.