Multi-Source Motion Estimation for Autonomous Accident Avoidance

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Autonomous vehicles in platooning scenarios may fail to react properly to accidents, leading to potential secondary collisions, and there is a need to effectively avoid high-risk surrounding objects.

Innovation Solution

A learning device that fuses information from images and communications to estimate the motions of surrounding objects using a combination of neural networks, including a combination of neural networks, including FC layers, to generate communication-based meta feature vectors and image-based meta feature vectors, and integrates them using a Recurrent Neural Network (RNN) to predict potential accidents and adjust driving parameters accordingly.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If autonomous vehicles use conventional platooning technology with basic collision avoidance, then fuel efficiency and road management improve, but the vehicles fail to react properly to accidents, leading to secondary collisions

Engineering Contradiction:
Improveaccident response capabilityVSAvoidsecondary collision risk
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The system performs preliminary actions by continuously estimating future positions and motions of surrounding objects using multiple prediction models before accidents occur. The learning device predicts potential collision scenarios in advance and prepares avoidance strategies, enabling the autonomous vehicle to react proactively rather than reactively to accident situations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system employs dynamic prediction models that continuously update estimates of surrounding objects' motions based on real-time sensor data and communication information. The prediction models adapt to changing conditions by adjusting trajectory predictions, velocity estimates, and acceleration patterns, allowing the vehicle to dynamically respond to evolving accident scenarios.

Inventive Principle:
Principle #15Dynamics

2Measurement precision

If autonomous vehicles rely on single-source information for decision making, then system complexity is reduced, but the ability to detect and avoid hazardous objects deteriorates

Engineering Contradiction:
Improvesurrounding object motion estimation accuracyVSAvoidinformation fusion system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The learning device merges multiple information sources including sensor data from cameras, LIDAR, and radar with communication information from other vehicles and infrastructure. By combining these diverse data streams through neural network-based fusion, the system achieves more accurate and robust estimation of surrounding objects' motions compared to single-source information systems.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system introduces intermediate processing layers including feature extraction modules and prediction models that mediate between raw sensor/communication data and final decision-making. These intermediary components transform multi-source information into standardized motion estimates, managing complexity by creating structured intermediate representations rather than directly processing all raw data.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Reliability

If the system uses multiple prediction models for comprehensive scene understanding, then accident detection accuracy improves, but processing time and computational load increase

Engineering Contradiction:
Improveaccident prediction reliabilityVSAvoiddecision-making response time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system segments the prediction task into multiple specialized models, each handling specific aspects such as pedestrian motion, vehicle trajectory, and object interaction. This segmentation allows parallel processing of different prediction streams, reducing overall computation time while maintaining comprehensive scene understanding through integration of specialized model outputs.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3690730B1Learning device for supporting safer autonomous driving without danger of accident by estimating motions of surrounding objects through fusion of information from multiple sources
Publication Date: 2025.12.10 STRADVISION
  • EP3690730B1 patent drawingFigure 1
  • EP3690730B1 patent drawingFigure 2A
  • EP3690730B1 patent drawingFigure 2B

AI summary

A learning method for supporting a safer autonomous driving through a fusion of information acquired from images and communications is provided. And the method includes steps of: (a) a learning device instructing a first neural network and a second neural network to generate an image-based feature map and a communication-based feature map by using a circumstance image and circumstance communication information; (b) the learning device instructing a third neural network to apply a third neural network operation to the image-based feature map and the communication-based feature map to generate an integrated feature map; (c) the learning device instructing a fourth neural network to apply a fourth neural network operation to the integrated feature map to generate estimated surrounding motion information; and (d) the learning device instructing a first loss layer to train parameters of the first to the fourth neural networks.