Trajectory Prediction Using Behavior Signals

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing trajectory prediction methods for objects in robotics and autonomous vehicles face challenges in accurately forecasting future paths, especially when dealing with human-driven vehicles or pedestrians, due to uncertainty in human behavior, leading to potential collisions.

Innovation Solution

A computer-implemented method that incorporates a behavior signal into a machine learning system to predict future trajectories, using input signals that include trajectory measurements and behavior classifications, such as aggressive or passive behavior, to condition the prediction process, allowing for improved accuracy by considering semantic behavior of objects like humans.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional machine learning methods are used for trajectory prediction without behavior signals, then the system complexity remains low, but the prediction accuracy deteriorates due to uncertainty in human behavior

Engineering Contradiction:
Improvetrajectory prediction accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The behavior classification is determined in advance as a behavior signal before the trajectory prediction process. This preliminary classification of human behavior (e.g., aggressive, defensive, cooperative) provides contextual information that guides the trajectory prediction, reducing uncertainty without requiring complex real-time analysis during prediction

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

A behavior signal acts as an intermediary between raw trajectory data and the machine learning prediction model. This intermediate representation captures semantic information about human behavior intentions, bridging the gap between observed movements and predicted future trajectories, thereby improving accuracy without directly increasing model complexity

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If behavior signals are incorporated into the machine learning system, then the trajectory prediction accuracy improves, but the data processing complexity increases

Engineering Contradiction:
Improvetrajectory prediction accuracyVSAvoiddata processing complexity
Core Design Contradiction:
Measurement precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The input data is segmented into distinct components: trajectory data and behavior signals. This segmentation allows the system to process each type of data independently through specialized modules, simplifying the overall processing complexity while maintaining high prediction accuracy through targeted analysis of each data type

Inventive Principle:
Principle #1Segmentation

3Reliability

If behavior classification is added as an input signal, then the system's ability to handle uncertain human behavior improves, but the information processing requirements increase

Engineering Contradiction:
Improvehandling of uncertain human behaviorVSAvoidinformation processing requirements
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The system transforms raw behavioral observations into classified behavior signals with discrete categories (e.g., aggressive, defensive, cooperative). This parameter transformation condenses complex behavioral information into manageable categories that the machine learning model can efficiently process, reducing information loss while improving reliability in handling uncertain human behavior

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4475089A1Device and method for determining a trajectory of an object
Publication Date: 2024.12.11 ROBERT BOSCH GMBH
  • EP4475089A1 patent drawingFigure 1
  • EP4475089A1 patent drawingFigure 2
  • EP4475089A1 patent drawingFigure 3

AI summary

Computer-implemented method for predicting an output signal (y) characterizing a future trajectory of an object, wherein the method comprises the steps of: • Obtaining an input signal (x) characterizing a measurement of a trajectory of the object; • Obtaining a behavior signal (b) characterizing a behavior classification of the trajectory or the future trajectory, e.g., classifying the trajectory as aggressive behavior or a passive behavior; • Determining the output signal (y) by providing the input signal (x) and the behavior signal (b) as input to a machine learning system (60), wherein the machine learning system (60) is configured to determine an output signal (y) characterizing a future trajectory an object based on the values of an input signal (x) characterizing a measurement of a trajectory of the object and a behavior signal (b) characterizing a behavior classification of the trajectory.