Automated Vehicle Predictive Control at Tire Handling Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive control systems for automated driving vehicles often violate physical constraints and handling limits, leading to safety issues such as tire saturation and collisions, especially in complex or unexpected driving scenarios, due to increased computational complexity and delays in encoding vehicle dynamics and road topology.
Innovation Solution
The implementation of a prediction model using Kalman filtering and non-linear model predictive control (NMPC) to estimate friction and optimize tire-force utilization, dynamic load transfer, and brake distribution, which adjusts parameters based on estimated road conditions and sideslip costs to stay within handling limits and improve traction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a prediction model uses static distribution of available forces and load constraints, then the control system is simple to implement, but tire saturation occurs and the vehicle violates handling limits
Solution Approach 1:
The patent transforms the static distribution of available forces into a dynamic distribution that adapts to changing road conditions and vehicle states. The prediction model now uses Kalman filtering to estimate friction coefficients in real-time and dynamically adjusts the distribution of braking forces among wheels based on estimated load transfer and actual traction potential, preventing tire saturation while maintaining handling limits compliance.
Solution Approach 2:
The system changes the parameter distribution from static to dynamic by introducing time-varying friction estimates and load transfer calculations. The prediction model continuously updates the available force distribution parameters based on measured vehicle states and estimated road conditions, allowing the control system to adapt to changing traction conditions without violating handling limits.
2Reliability
If the prediction model encodes vehicle dynamics, physical constraints, road topology, and costs, then safety is improved by reducing approximation errors, but computational complexity and costs increase
Solution Approach 1:
The patent segments the computational tasks by separating the friction estimation (using Kalman filtering) from the optimal control optimization (using NMPC). This division allows the system to handle complex vehicle dynamics and constraints through modular computation, reducing overall computational complexity while maintaining safety through accurate friction estimates that guide the optimization process.
Solution Approach 2:
The system performs preliminary friction estimation using Kalman filtering before executing the computationally intensive NMPC optimization. By pre-establishing accurate friction coefficients and load transfer estimates, the system reduces the computational burden during the optimization phase, as the NMPC can focus on finding optimal control inputs rather than estimating fundamental parameters.
3Reliability
If the vehicle performs online encoding to calculate and reduce costs for projected trajectories, then collision probability is reduced, but delays occur that increase collision probabilities in sudden danger scenarios
Solution Approach 1:
The system performs preliminary friction estimation and parameter adjustment using Kalman filtering before the NMPC optimization is executed. This pre-computation of essential parameters (friction coefficients, load transfer estimates) accelerates the overall control loop by reducing the computational burden during the trajectory optimization phase, enabling faster reaction to sudden dangers while maintaining collision prevention capabilities.
Solution Approach 2:
The Kalman filter provides continuous feedback on friction estimates and vehicle states, allowing the NMPC to rapidly adjust control commands based on current traction conditions. This feedback mechanism enables the system to respond quickly to changing road conditions and sudden dangers, reducing computation delays by using updated parameter estimates rather than requiring full re-encoding of vehicle dynamics.
4Reliability
If the prediction model uses dynamic load transfer and brake distributions, then tire-force utilization is optimized and handling limits are respected, but computational requirements increase
Solution Approach 1:
The patent segments the computational tasks by separating friction estimation (Kalman filtering) from optimal control optimization (NMPC). This modular approach allows dynamic load transfer and brake distribution calculations to be performed efficiently by leveraging pre-estimated friction parameters, reducing overall computational requirements while maintaining optimized tire-force utilization.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances safety and comfort by optimizing tire-force utilization, reducing traction loss, and avoiding collisions by leveraging available traction potential and tire friction, while also reducing computational complexity and delays.
Implementation Method 1
The prediction system adjusts a prediction model by implementing Kalman filtering that estimates friction involving a projected trajectory from an automated driving system (ADS)
Implementation Method 2
The prediction model generates vehicle dynamics using a load transfer and a brake distribution, the vehicle dynamics being associated with estimated road conditions and the handling limits
Implementation Method 3
The prediction model optimizes tire-force utilization (e.g., saturation) through the load transfer and brake distribution according to estimated road conditions
Data Source
AI summary
System, methods, and other embodiments described herein relate to adjusting a prediction model for control at handling limits associated with a projected trajectory during automated driving. In one embodiment, a method includes adjusting parameters of a prediction model using friction estimates and sideslip costs associated with a projected trajectory of a vehicle, the friction estimates being derived from Kalman filtering. The method also includes scaling, using the prediction model, handling limits of the vehicle for the projected trajectory according to a friction circle. The method also includes generating, by the prediction model, vehicle dynamics using a load transfer and a brake distribution, the vehicle dynamics being associated with estimated road conditions and the handling limits. The method also includes outputting, by the prediction model using the vehicle dynamics, a driving command to the vehicle for the projected trajectory.