Human Behavior Prediction Models for Complex Road Interactions

Solution Overview

Problem

Current systems, such as autonomous driving vehicles, struggle to accurately predict human behavior beyond motion vectors, leading to inferior results in anticipating pedestrian, cyclist, and motorist actions, especially in complex scenarios like crowded areas.

Innovation Solution

A system that uses a computing device to process images and video segments, generating stimulus data, aggregating user responses to create statistical models, and applying these models to predict human behavior, incorporating supervised learning algorithms like random forests and neural networks to improve prediction accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If motion vector extrapolation methods are used to predict human behavior, then the prediction process is simple and fast, but the prediction accuracy is inferior

Engineering Contradiction:

Improveprediction accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary training system that collects human observer responses to various scenarios and uses these responses to train machine learning models. This trained model then serves as the mediator between raw video data and behavior prediction, providing accurate predictions without requiring complex manual analysis of each scenario. The intermediary trained model encapsulates human-like reasoning capabilities.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary action by training the prediction model in advance using a comprehensive dataset of human observer responses to various scenarios. This pre-training phase allows the model to learn complex human behavior patterns beforehand, so that during actual operation, predictions can be made quickly and accurately without real-time complex processing.

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If human observer responses are collected and aggregated to train prediction models, then prediction accuracy improves, but data processing time and computational resources increase

Engineering Contradiction:

Improvebehavior prediction accuracyVSAvoiddata processing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary action by training the prediction model in advance using a comprehensive dataset of human observer responses to various scenarios. This pre-training phase allows the model to learn complex human behavior patterns beforehand, so that during actual operation, predictions can be made quickly and accurately without real-time complex processing.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements continuous refinement of the prediction model through ongoing collection and processing of human observer responses. The model is continuously trained and updated, allowing the system to maintain high prediction accuracy while the processing workload is distributed over time rather than concentrated during critical prediction moments.

Inventive Principle:

Principle #20Continuity of useful action

3Measurement precision

If complex scenarios like crowded areas are analyzed in detail, then prediction accuracy improves, but processing speed decreases

Engineering Contradiction:

Improveprediction accuracy in complex scenariosVSAvoidprocessing speed

Core Design Contradiction:

Measurement precisionVSSpeed

Solution Approach 1:

The patent replaces mechanical analysis methods with machine learning-based prediction. Instead of using complex algorithms to manually analyze each scenario in real-time, the system uses a trained neural network model that has learned from extensive human observer data. This substitution allows the system to maintain high prediction accuracy in complex scenarios while achieving real-time processing speeds.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs preliminary action by training the prediction model in advance using a comprehensive dataset of human observer responses to various scenarios. This pre-training phase allows the model to learn complex human behavior patterns beforehand, so that during actual operation, predictions can be made quickly and accurately without real-time complex processing.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11981352B2System and method of predicting human interaction with vehicles

Publication Date: 2024.05.14 PERCEPTIVE AUTOMATA INC

AI summary

Systems and methods for predicting user interaction with vehicles. A computing device receives an image and a video segment of a road scene, the first at least one of an image and a video segment being taken from a perspective of a participant in the road scene and then generates stimulus data based on the image and the video segment. Stimulus data is transmitted to a user interface and response data is received, which includes at least one of an action and a likelihood of the action corresponding to another participant in the road scene. The computing device aggregates a subset of the plurality of response data to form statistical data and a model is created based on the statistical data. The model is applied to another image or video segment and a prediction of user behavior in the another image or video segment is generated.