Route Planning Model Generation Using Multi-Site Historical Data

Solution Overview

Problem

Autonomous parking systems face challenges in generating accurate route planning models for new sites due to the need for extensive training data, which increases user experience costs and computational overhead.

Innovation Solution

A method and apparatus for generating a target route planning model using a site optimization object and a first training optimization object, where the target route planning model is determined based on historical route data from multiple sites, reducing the requirement for large amounts of target site-specific data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If extensive training data is collected for each new site, then the accuracy of route planning model is improved, but the user experience cost and computational overhead increase

Engineering Contradiction:

Improveroute planning model accuracyVSAvoiduser experience cost

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary training on a first route planning model using historical route data from multiple sites before deployment. This pre-training establishes a foundational model that can be quickly adapted to new sites with minimal additional training data, thereby improving route planning accuracy without requiring extensive on-site data collection that would increase user experience cost

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The first route planning model is trained on diverse historical route data from multiple different sites, creating a universal model that can generalize across various environments. This multi-functional model serves as a foundation for generating target route planning models for new sites, reducing the need for site-specific extensive training while maintaining accuracy

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Measurement precision

If extensive training data is collected for each new site, then the accuracy of route planning model is improved, but the computational overhead increases

Engineering Contradiction:

Improveroute planning model accuracyVSAvoidcomputational overhead

Core Design Contradiction:

Measurement precisionVSPower

Solution Approach 1:

The system performs preliminary training on a first route planning model using historical route data from multiple sites before deployment. This pre-training establishes a foundational model that can be quickly adapted to new sites with minimal additional training data, thereby improving route planning accuracy without requiring extensive on-site data collection that would increase user experience cost

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system changes the training parameters by using a pre-trained first route planning model as the basis for generating target route planning models. Instead of training from scratch with extensive data, the system adjusts the existing model using site optimization objects and target route data sets, significantly reducing computational overhead while maintaining model accuracy

Inventive Principle:

Principle #35Parameter changes

3Loss of time

If site-specific training data is reduced, then the user experience cost is reduced, but the accuracy of route planning model may deteriorate

Engineering Contradiction:

Improveuser experience costVSAvoidroute planning model accuracy

Core Design Contradiction:

Loss of timeVSMeasurement precision

Solution Approach 1:

The first route planning model is trained on diverse historical route data from multiple different sites, creating a universal model that can generalize across various environments. This multi-functional model serves as a foundation for generating target route planning models for new sites, reducing the need for site-specific extensive training while maintaining accuracy

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system introduces a site optimization object as an intermediary that bridges the gap between the pre-trained first route planning model and the target site requirements. This intermediary enables effective adaptation to new sites with minimal training data by optimizing the model parameters specifically for the target environment while leveraging the general knowledge from historical data

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11698262B2Method and apparatus for generating route planning model, and storage medium

Publication Date: 2023.07.11 BEIJING BAIDU NETCOM SCI & TECH CO LTD

AI summary

A method and an apparatus for generating a route planning model and a storage medium are provided. The method for generating a route planning model includes: obtaining a target route data set associated with a target site; and determining a target route planning model of the target site with a site optimization object corresponding to the target site, based on the target route data set and a first route planning model; wherein the first route planning model is determined based on a set of historical route data through at least a first training, the set of historical route data being associated with a plurality of sites different from the target site and a first training optimization object for the first training corresponding to the plurality of sites.