Context-Based Traffic Prediction Using Historical Indicators

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current driver assistance systems fail to accurately predict the future behavior of target vehicles due to their reliance on only the current traffic situation, leading to delayed reactions and reduced prediction quality, as they do not consider the historical context and previous behaviors of traffic participants.

Innovation Solution

The system employs a combination of context-based and physical predictions, incorporating direct and indirect indicators, including historical data about the target vehicle's previous behavior and its relations with other traffic participants and infrastructure, to improve the accuracy of future behavior estimation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If only current traffic situation is considered for prediction, then the system complexity is reduced, but the prediction accuracy deteriorates

Engineering Contradiction:
Improveprediction accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary analysis of historical behavior data and establishes behavior models in advance. Indirect indicators are pre-calculated from historical data, and behavior probabilities are estimated before critical situations occur, enabling faster and more accurate real-time predictions without excessive computational complexity during critical moments

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces indirect indicators as intermediary elements that mediate between raw historical data and prediction outcomes. These indirect indicators (such as behavior patterns, contextual features) serve as intermediate representations that capture essential information from historical data without requiring the system to process all raw data directly, thus improving accuracy while managing complexity

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If historical data is incorporated into prediction, then the prediction accuracy is improved, but the processing time increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system extracts only the most relevant features and indirect indicators from historical data that are necessary for prediction, rather than processing the entire historical dataset. This selective extraction of critical information maintains prediction accuracy while significantly reducing processing time by focusing on essential behavioral patterns and contextual features

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial action by considering only the most relevant historical behaviors and indicators needed for the current prediction context, rather than analyzing all possible historical data. This partial processing approach provides sufficient accuracy for safety-critical decisions while minimizing processing time delays

Inventive Principle:
Principle #16Partial or excessive action

3Reliability

If context-based prediction is used without historical context, then the system responds faster to current situations, but the reliability of prediction deteriorates

Engineering Contradiction:
Improveprediction reliabilityVSAvoidprediction model complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The prediction model is segmented into multiple independent components: direct indicators from current sensor data, indirect indicators from historical data, and behavior probability calculations. Each segment processes specific types of information independently, then combines results to produce the final prediction. This segmentation improves reliability by ensuring no single component fails the entire system while managing overall model complexity through modular design

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system incorporates feedback mechanisms where historical behavior outcomes feed back into updating behavior models and indirect indicators. Past prediction accuracy and actual behavior outcomes are used to refine the behavior models, improving future prediction reliability. This feedback loop enhances reliability by continuously learning from historical data while maintaining manageable complexity through iterative refinement rather than complex upfront modeling

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP2923911B1A method and system for predicting movement behavior of a target traffic object
Publication Date: 2019.03.13 HONDA RES INST EUROPE
  • EP2923911B1 patent drawingFigure 1
  • EP2923911B1 patent drawingFigure 2~2b
  • EP2923911B1 patent drawingFigure 3~3b

AI summary

The invention regards a method for computationally predicting future movement behavior of a target object and program comprising the steps of producing sensor data by at least one sensor physically sensing the environment of a host vehicle, computing a plurality of movement behavior alternatives of the target object sensed by the sensors, by predicting movement behaviors of the target object applying a context based prediction step using at least one indirect indicator and/or indicator combinations derived from sensor data, wherein in said context based prediction step a probability that the target object will execute a movement behavior at a time is estimated. A future position of the target object is estimated and a signal representing the estimated future position is outputted. In the context based prediction step at least one history indicator for at least one movement behavior alternative is generated for a current point in time using at least one indicator value of an indirect indicator at a point in time in the past.