Probabilistic PDZ Availability Estimation for Autonomous Routing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Autonomous vehicles face challenges in efficiently locating available pick-up or drop-off zones (PDZs) due to the lack of real-time information, leading to unnecessary time consumption and inefficiencies in ride or delivery services.

Innovation Solution

A probabilistic model is developed to estimate PDZ availability using historical data from various sources, including user reports, vehicle logs, sensor data, and traffic information, which is trained using machine learning algorithms and refined in real-time or periodically to generate routes that optimize the likelihood of finding available PDZs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If autonomous vehicles search for available PDZs without real-time information, then they can operate with simple systems, but they consume unnecessary time and reduce service efficiency

Engineering Contradiction:

Improveservice delivery efficiencyVSAvoidtime spent searching for PDZs

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by collecting historical PDZ availability data, training machine learning models, and generating probabilistic availability estimates before vehicles need to locate PDZs. This advance preparation enables vehicles to efficiently plan routes to likely available PDZs without time-consuming searches

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements feedback by continuously collecting real PDZ availability observations from vehicles and users, using this data to train and refine machine learning models, and updating probabilistic availability estimates. This closed-loop feedback improves prediction accuracy over time, reducing search time and improving service efficiency

Inventive Principle:

Principle #23Feedback

2Loss of time

If a probabilistic model with multiple data sources is used to estimate PDZ availability, then route planning efficiency improves, but system complexity increases

Engineering Contradiction:

Improvetime spent searching for PDZsVSAvoidcomplexity of probabilistic model system

Core Design Contradiction:

Loss of timeVSDevice complexity

Solution Approach 1:

The system achieves universality by using a single machine learning model framework that processes multiple data sources (historical availability data, real-time observations, traffic information, weather data) through unified algorithms. This multi-functional approach handles diverse inputs while maintaining a cohesive system architecture, balancing complexity with comprehensive data utilization

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS11300419B2Pick-up/drop-off zone availability estimation using probabilistic model

Publication Date: 2022.04.12 UBER TECHNOLOGIES INC

AI summary

Aspects of the present disclosure include systems, methods, and devices to provide estimations of vehicular pick-up/drop-off zone (PDZ) availability. A request for vehicular PDZ availability at a location is received from a vehicular autonomy system of a vehicle. The request specifies an estimated time of arrival at the location. The PDZ availability at the location at the estimated time of arrival is estimated using a probabilistic model. A response to the request is generated based on the estimated PDZ availability. The response indicates the estimated PDZ availability. The response is transmitted to the vehicular autonomy system responsive to the request.