Unified Destination Entry Tiling for In-Car Speech Context Loading

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing mapping systems face challenges in providing location-dependent information efficiently, as they often rely on large monolithic data structures that consume significant memory, and users interacting with map data through speech interfaces lack sufficient context data for accurate responses.

Innovation Solution

The system utilizes context data, referred to as 'unified destination entry data' (UDE data), which is partitioned into tiles that are dynamically loaded based on the user's location. This approach includes loading the local tile and neighborhood tiles into memory, and decomposing remote tiles into overlay tiles to optimize memory usage.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If context data is partitioned into tiles and dynamically loaded based on user location, then memory usage is optimized, but the complexity of data management increases

Engineering Contradiction:
Improvememory usageVSAvoiddata management complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent partitions context data into discrete geographic tiles that can be independently loaded and managed. Each tile contains context information for a specific geographic region, allowing the system to load only the tiles relevant to the user's current location rather than loading all context data into memory. This segmentation directly reduces memory usage while the tile-based structure provides a manageable framework for data organization.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically loads and unloads context tiles based on the user's current location and movement. As the user moves between geographic regions, the system automatically loads new tiles and unloads previously loaded tiles that are no longer needed. This dynamic approach optimizes memory usage by ensuring that only currently relevant context data resides in memory, while the underlying data structure remains organized and easily manageable through location-based criteria.

Inventive Principle:
Principle #15Dynamics

2Quantity of substance

If remote tiles are decomposed into overlay tiles, then memory usage is further optimized, but the device complexity increases

Engineering Contradiction:
Improvememory usageVSAvoidtile structure complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent further segments remote tiles by decomposing them into multiple overlay tiles, each representing a subset of the original remote tile's geographic area. This allows the system to load only specific portions of remote tiles that are relevant to the user's location or area of interest, rather than loading entire remote tiles. This finer-grained segmentation reduces memory usage while maintaining an organized hierarchical structure where overlay tiles can be managed independently.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The overlay tile structure enables the system to provide different levels of detail and information density for different geographic regions. Local tiles near the user's current location can be fully loaded with complete context information, while remote tiles can be represented by selective overlay tiles containing only the most relevant information. This local quality approach optimizes memory usage by allocating memory resources according to the importance and proximity of different geographic regions.

Inventive Principle:
Principle #3Local quality

3Quantity of substance

If only local tile and neighborhood tiles are loaded, then memory usage is optimized, but the accuracy of speech recognition may deteriorate

Engineering Contradiction:
Improvememory usageVSAvoidspeech recognition accuracy
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The system loads not only the strictly necessary local tile but also additional neighborhood tiles that surround the user's current location. This partial extension beyond the minimum required data ensures that context information is available for speech recognition even when the user's location changes slightly or when referring to nearby locations. The neighborhood tiles provide a buffer zone of context data that maintains speech recognition accuracy while still limiting overall memory usage compared to loading all possible context data.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4538644A1Tiling based approach for unified destination entry
Publication Date: 2025.04.16 CERENCE OPERATING CO
  • EP4538644A1 patent drawingFigure 1
  • EP4538644A1 patent drawingFigure 2
  • EP4538644A1 patent drawing

AI summary

In an infotainment system of a vehicle, a mass storage unit stores map data and context data. The map data comprises geographic information about a geographic area around the vehicle and the context data comprises location-dependent context information about the geographic area for use when engaging in speech interaction with an occupant of the vehicle. The context data includes subsets of context data, each of which is pertinent to a different geographic area. A memory manager copies these subsets of context data into a dynamic memory based on the vehicle's movement.