Autonomous Vehicle Path Control for Human-Like Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle path planning methods, such as sampling-based approaches, generate paths that may feel unnatural to passengers due to their opportunistic and exploratory nature, diverging from human driving behaviors which are result-oriented and adaptable.
Innovation Solution
Implementing a driving objective-based path planning system that mimics human driving by formulating objectives like maintaining speed, headway, and lane position, using probabilistic control functions to balance multiple objectives and adapt to driving styles and environmental uncertainties, integrating low-rate motion planning with high-rate vehicle control for adaptive and comfortable navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sampling-based path planning methods are used, then fast machine calculation is achieved, but the generated path feels unnatural to passengers
Solution Approach 1:
The patent copies human driver behavior patterns into the autonomous vehicle's decision-making system. By observing and replicating how human drivers naturally navigate, prioritize objectives, and adapt to situations, the system generates paths that feel natural to passengers while maintaining computational efficiency through learned behavioral patterns rather than exhaustive sampling
Solution Approach 2:
The patent changes the fundamental parameters of path planning from probabilistic sampling to deterministic objective-based control. By shifting from random exploration to goal-directed navigation with adjustable parameters for different driving styles and situations, the system achieves both computational speed and natural-looking paths
2Adaptability or versatility
If multiple control objectives are balanced using probabilistic control functions, then adaptability to different driving styles and environmental uncertainties is improved, but computational complexity increases
Solution Approach 1:
The patent segments the control system into distinct modular components: objective formulation module, probabilistic control function module, and parameter adjustment module. Each module handles specific aspects of adaptive control independently, making the overall complex system manageable and allowing selective activation of different control strategies based on situational needs
Solution Approach 2:
The patent implements dynamic adjustment of control parameters and objective priorities based on real-time environmental conditions and detected driving styles. The system continuously adapts its behavior by modifying parameter values rather than restructuring the entire control architecture, maintaining versatility while controlling complexity through parameter-based flexibility
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
Controller of a vehicle uses control functions to transition the current state of the vehicle into a target state. A control function is probabilistic to output a parametric probability distribution over the target state defined by a first moment and at least one higher order moment. The controller submits the current state into at least a subset of control functions consistent with the next driving decision to produce a subset of parametric probability distributions over the target state, combines the subset of parametric probability distributions to produce a joint parametric probability distribution of the target state, and determines the control command based on the first moment and at least one higher order moment of the joint parametric probability distribution of the target state.