Vehicle MPC Parameter Updating With Simulation-Learned Controls

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current model predictive controllers (MPC) for autonomous vehicles rely on fixed operational, external, and control parameters, which do not account for dynamic changes in vehicle and environmental conditions, leading to suboptimal performance.

Innovation Solution

Implementing a trained machine learning vehicle performance circuit to dynamically update operational parameters, such as vehicle and external parameters, and control gains, by simulating various scenarios to determine optimal values for path tracking, corridor keeping, and collision avoidance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If fixed operational parameters are used in model predictive controllers, then the controller structure is simple and easy to implement, but the controller cannot adapt to dynamic changes in vehicle and environmental conditions, leading to suboptimal performance

Engineering Contradiction:

Improveadaptability to dynamic changesVSAvoidcontroller structure complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies dynamics by transitioning from fixed parameters to dynamically adjustable parameters. A machine learning circuit continuously learns and updates operational parameters (vehicle parameters, external parameters, and controls parameters) based on real-time sensor data and simulated scenarios, enabling the MPC to adapt its behavior to changing conditions while maintaining a relatively simple controller structure.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent uses preliminary action by pre-training the machine learning circuit through extensive simulations across a wide range of operational conditions before actual vehicle deployment. This pre-training allows the system to have optimized parameters ready in advance for various scenarios, reducing the computational burden during real-time operation and enabling fast adaptation without complex real-time optimization.

Inventive Principle:

Principle #10Preliminary action

2Reliability

If learned operational parameters are used to adapt to dynamic changes, then the controller performance is improved, but the system complexity and computational requirements increase

Engineering Contradiction:

Improvecontroller performanceVSAvoidsystem complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies copying by creating a virtual replica of the vehicle and its operating environment through high-fidelity simulations. The machine learning circuit is trained on this virtual copy under diverse and extreme conditions that would be dangerous or impractical to test in the real vehicle. This allows the system to achieve high reliability through extensive virtual testing without proportionally increasing physical system complexity.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces complex real-time mechanical adjustment systems with a software-based machine learning approach. Instead of physically reconfiguring vehicle components based on conditions, the system uses learned parameters to dynamically adjust control decisions, achieving adaptive performance through computational methods rather than mechanical complexity.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If extensive simulations are conducted to train the machine learning circuit, then the accuracy of learned parameters is improved, but the training time and computational resources increase

Engineering Contradiction:

Improveaccuracy of learned parametersVSAvoidtraining time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary action by conducting comprehensive simulations and training the machine learning circuit before actual vehicle deployment. Extensive simulations are run offline to cover a wide range of operational scenarios, weather conditions, and vehicle states, allowing the system to accumulate high-accuracy learned parameters in advance. This upfront investment in training time enables fast, accurate parameter retrieval during real-time operation.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent applies dynamics by implementing a continuous learning mechanism that updates parameters in real-time based on actual vehicle operation data. The machine learning circuit adapts to new conditions dynamically, refining its accuracy over time without requiring complete retraining. This allows the system to maintain high precision while reducing the computational burden compared to static, pre-computed approaches.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS12459497B2Systems and methods for updating the parameters of a model predictive controller with learned controls parameters generated using simulations and machine learning

Publication Date: 2025.11.04 TOYOTA JIDOSHA KK

AI summary

A computer implemented method for determining optimal values for controls parameters for a model predictive controller for controlling a vehicle can receive from a data store or a graphical user interface, ranges for one or more operational parameters. The computer implemented method can determine optimum values for controls parameters by simulating a vehicle operation across the ranges of the one or more operational parameters by solving a vehicle control problem and determining an output of the vehicle control problem based on a result for the simulated vehicle operation. A vehicle can include a processing component configured to adjust a control input for an actuator of the vehicle according to a control algorithm and based on the optimum values of the controls parameter as determined by the computer implemented method.