Dispatching Enhanced Sensor Vehicles for Autonomous Driving Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional transportation systems face challenges in accuracy, efficiency, and flexibility when generating precise learning data for computer models, often relying on synthetic training data or heuristic models that result in inaccurate and imprecise machine learning models, particularly for autonomous vehicle driving systems.

Innovation Solution

A transportation matching system that dynamically generates learning data by intelligently dispatching enhanced sensor provider vehicles using data collection values, which are determined based on route features, geofencing, and historical data to prioritize data collection, thereby improving the accuracy and efficiency of data gathering for computer-implemented models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If conventional transportation systems use synthetic training data or heuristic models, then device complexity is reduced, but manufacturing precision (accuracy of machine learning models) deteriorates

Engineering Contradiction:

Improvesystem complexityVSAvoidaccuracy of machine learning models

Core Design Contradiction:

Device complexityVSManufacturing precision

Solution Approach 1:

The system performs preliminary data collection by dispatching enhanced sensor provider vehicles to collect real-world sensory data before training machine learning models. This advance data gathering ensures high-quality training data is available, improving model accuracy without requiring complex synthetic data generation processes during model training.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces enhanced sensor provider vehicles as intermediaries between the transportation system and the machine learning model training process. These vehicles equip specialized sensors to collect precise real-world data, serving as a bridge that provides accurate training data without requiring the system to build complex synthetic data generation capabilities.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Manufacturing precision

If enhanced sensor provider vehicles are dispatched to collect precise training data, then manufacturing precision (accuracy of machine learning models) is improved, but productivity (efficiency of servicing transportation requests) deteriorates

Engineering Contradiction:

Improveaccuracy of machine learning modelsVSAvoidefficiency of servicing transportation requests

Core Design Contradiction:

Manufacturing precisionVSProductivity

Solution Approach 1:

The enhanced sensor provider vehicles perform multiple functions simultaneously: they service transportation requests while also collecting training data for machine learning models. This multi-functionality ensures that the system maintains productivity by not dedicating separate resources solely to data collection, while still achieving high model accuracy through real-world data gathering.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system ensures continuous data collection by integrating it into the ongoing transportation operations. Enhanced sensor provider vehicles continuously gather training data during their regular transportation duties, eliminating idle data collection phases and maintaining uninterrupted productivity while improving model accuracy over time.

Inventive Principle:

Principle #20Continuity of useful action

3Manufacturing precision

If real-world sensory data is collected from enhanced sensor provider vehicles, then manufacturing precision (accuracy of machine learning models) is improved, but loss of time (data collection time) increases

Engineering Contradiction:

Improveaccuracy of machine learning modelsVSAvoiddata collection time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The system collects real-world sensory data in advance during regular transportation operations before model training is needed. This preliminary data collection during routine operations eliminates the need for separate dedicated data collection trips, reducing time loss while ensuring sufficient training data is available for accurate model training.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20220327933A1Intelligently generating computer model learning data by dispatching enhanced sensory vehicles utilizing data collection values

Publication Date: 2022.10.13 LYFT INC

AI summary

The present disclosure relates to systems, non-transitory computer-readable media, and methods for determining data collection values for enhanced sensor provider vehicles in a provider vehicle pool and utilizing the data collection values to select and dispatch a provider vehicle for a transportation request. In particular, in one or more embodiments, the disclosed systems determine data collection values based on various route features for a particular enhanced sensor provider vehicle. Additionally, in one or more embodiments, the disclosed systems can utilize the data collection values in conjunction with a device matching algorithm and transportation values to select and dispatch a provider vehicle. Further, in some embodiments, the disclosed systems utilize collected sensory data to train one or more computer implemented machine learning models, such as an autonomous vehicle driving model.