Autonomous Driving Trajectory Validation with Redundant Sensor Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing solutions for determining vehicle trajectories in autonomous vehicles lack reliability and robustness, especially in complex and dynamic environments, which can lead to errors and safety issues, particularly for vehicles at Level 2 and above as defined by the SAE J3016 standard.

Innovation Solution

A method involving the creation of multiple environmental models, including a primary model using all available sensor data and secondary models using subsets of sensor data, where the primary trajectory is validated against secondary models to ensure reliability and robustness, utilizing sensor fusion and machine learning for accurate navigation and collision checking.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a single environmental model is used for trajectory determination, then the system is simpler and faster, but reliability and robustness are insufficient in complex and dynamic environments

Engineering Contradiction:
Improvetrajectory determination reliabilityVSAvoidenvironmental model architecture complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the environmental model architecture into a primary environmental model and multiple secondary environmental models. Each model processes sensor data independently to generate trajectory predictions, allowing the system to evaluate multiple hypotheses simultaneously. This segmentation enables reliable trajectory determination by comparing results across models while maintaining manageable complexity through modular design.

Inventive Principle:
Principle #1Segmentation

2Reliability

If multiple environmental models are used to improve reliability, then robustness increases, but computational complexity and processing time increase

Engineering Contradiction:
Improvetrajectory determination reliabilityVSAvoidtrajectory computation speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies preliminary action by having secondary environmental models pre-process sensor data and generate trajectory predictions in parallel with the primary model. This allows the system to prepare multiple trajectory hypotheses in advance, so when validation is needed, the computations are already complete or near-complete, reducing overall processing time while maintaining high reliability.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If hardware redundancy is implemented to improve safety, then fault tolerance increases, but the system does not fulfill SOTIF architectural targets for Level 2 and above automation

Engineering Contradiction:
Improvesafety under fault conditionsVSAvoidsafety under non-fault conditions
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent changes the parameter of redundancy from hardware duplication to algorithmic diversity. Instead of using identical hardware components in parallel, the system employs multiple secondary environmental models with different sensor subsets and processing approaches. This allows the system to address SOTIF requirements by adapting to various environmental conditions and sensor failures while maintaining computational efficiency and architectural flexibility suitable for Level 2 and above automation.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4523990A1Redundant architecture for automated or autonomous driving
Publication Date: 2025.03.19 AIMOTIVE KFT
  • EP4523990A1 patent drawingFigure 1
  • EP4523990A1 patent drawingFigure 2
  • EP4523990A1 patent drawingFigure 3

AI summary

A method for determining trajectories for vehicles includes accessing a plurality of sensors (106) of a vehicle to define a plurality of environmental models, including a primary environmental model (112) and a plurality of secondary environmental models (114), wherein the primary environmental model uses sensor data of the plurality of sensors, and each secondary environmental model uses sensor data of a subset of the plurality of sensors, computing a primary trajectory (116) for the vehicle using the primary environmental model, and validating (120) the primary trajectory against each one of the plurality of secondary environmental models. Furthermore, a corresponding device and vehicle are defined.