Driver Assistance Steering Control With Intent-Predictive MPC

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing driver assistance systems fail to account for the driver's intentions, leading to reduced situational awareness and increased risk during transitions from automated to manual steering control, and do not effectively handle unseen driving situations.

Innovation Solution

A hybrid controller combining a neural network-based driver model with model predictive control (MPC) to predict and adjust vehicle controls, incorporating haptic feedback for improved driver collaboration and comfort.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If a driver assistance system performs steering tasks autonomously (SAE level 2 or lower), then the driver's workload is reduced and road safety is improved, but the driver's situational awareness deteriorates and the risk of unsafe take-over increases during transitions from assisted to manual steering control

Engineering Contradiction:
Improvedriver workloadVSAvoidsituational awareness
Core Design Contradiction:
Ease of operationVSReliability

Solution Approach 1:

The system performs preliminary actions by predicting the driver's intended steering torque using a neural network model before the driver actually applies it. The predicted torque is used to pre-adjust the electric power steering system, so when the driver takes over control, the steering wheel is already in a state that aligns with the driver's intention, preventing sudden conflicts and maintaining situational awareness.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements feedback by continuously monitoring the driver's steering interactions and using this information to update the neural network model's prediction of driver intent. The system compares the predicted torque with actual driver input and adjusts accordingly, creating a closed-loop feedback mechanism that maintains alignment between the assist system and driver intentions during transitions.

Inventive Principle:
Principle #23Feedback

2Device complexity

If a driver assistance system uses traditional control methods without driver behavior prediction, then the system structure remains simple, but the system cannot anticipate driver intentions and may conflict with driver actions

Engineering Contradiction:
Improvecontrol system structureVSAvoiddriver intention alignment
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The neural network model acts as an intermediary between the driver and the electric power steering system. Instead of directly controlling steering torque based on simple sensor feedback, the system uses the neural network to interpret and predict driver intentions, then uses this prediction to mediate the control signals sent to the steering system, enabling better alignment without requiring complete system redesign.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Adaptability or versatility

If a neural network model is trained on historical driving data to predict driver behavior, then the system can anticipate driver intentions and reduce conflicts, but the system performance deteriorates when encountering unseen driving situations not covered in training data

Engineering Contradiction:
Improvedriver behavior predictionVSAvoidperformance on unseen situations
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The system implements dynamics by making the neural network model adaptive and updatable. Rather than using a static model trained once on historical data, the system continuously learns from new driving interactions, adjusting its predictions based on evolving driver behavior patterns. This dynamic adaptation allows the system to maintain reliability when encountering new driving situations by incorporating real-time learning.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4699887A1Methods and systems for driver assistance
Publication Date: 2026.02.25 TOYOTA JIDOSHA KK
  • EP4699887A1 patent drawingFigure 1~2
  • EP4699887A1 patent drawingFigure 3
  • EP4699887A1 patent drawingFigure 4

AI summary

A computer-implemented method for assisting a driver of a vehicle comprising: (S10) measuring environment data representative of a driving situation of a vehicle on a road; (S20) generating predicted control variables based on the environment data by using a neural network-based model that is trained to learn the behavior of the driver; (S30) generating optimized control variables based on the predicted control variables, wherein the predicted control variables are inputted into a model predictive controller as reference targets over a receding horizon; and (S40) applying the optimized control variables to the vehicle controls.