Adaptive Vehicle Control Using Confidence-Weighted Trajectory Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous and semi-autonomous vehicles face challenges in motion planning due to the non-convexity of the motion-planning problem, leading to locally optimal solutions that may not be globally optimal, and existing sampling-based methods can result in unnatural paths, necessitating improved path planning and control systems.
Innovation Solution
An integrated architecture combining a low-rate, long-term sampling-based motion planning algorithm with a high-rate, highly reactive optimization-based predictive vehicle controller, utilizing probabilistic methods and particle filtering to compute confidence bounds and adapt the tracking cost function, allowing for automatic tuning of competing objectives such as safety and comfort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sampling-based motion planning methods are used, then computational speed is improved, but path naturalness deteriorates
Solution Approach 1:
The patent divides the motion planning task into two distinct layers: a low-rate sampling-based planner that generates candidate trajectories, and a high-rate optimization-based controller that refines these trajectories. This segmentation allows each layer to specialize - the planner explores the state space efficiently while the controller ensures natural and smooth paths, thus resolving the contradiction between computational speed and path naturalness.
Solution Approach 2:
The patent introduces a confidence-based cost function as an intermediary mechanism that bridges the sampling-based planner and the optimization-based controller. The confidence bounds computed by the planner are used to dynamically weight the tracking cost in the controller's objective function, enabling smooth transitions and natural paths while maintaining computational efficiency.
2Ease of manufacture
If deterministic sampling methods are used, then computational simplicity is improved, but path quality deteriorates
Solution Approach 1:
The patent employs dynamic confidence bounds that are computed based on the uncertainty of the motion planner's trajectory predictions. These confidence bounds are not static but adapt dynamically during execution, allowing the system to maintain simple deterministic sampling while achieving high path quality through adaptive weighting in the cost function.
Solution Approach 2:
The patent changes the parameters of the cost function dynamically based on the confidence bounds computed by the sampling-based planner. By adjusting the tracking cost weights according to the planner's confidence in different regions of the state space, the system achieves high path quality without complicating the sampling method itself.
3Manufacturing precision
If probabilistic sampling methods are used, then path quality is improved, but computational complexity increases
Solution Approach 1:
The patent applies probabilistic sampling selectively rather than throughout the entire control pipeline. The sampling-based planner uses probabilistic methods to generate candidate trajectories and compute confidence bounds, while the optimization-based controller uses these bounds to guide deterministic refinement. This partial application of probabilistic methods achieves high path quality without the full computational burden of purely probabilistic approaches.
4Speed
If high-rate control is implemented, then responsiveness is improved, but computational load increases
Solution Approach 1:
The patent segments the control system into two operational rates: a low-rate sampling-based planner that runs at lower frequency to generate candidate trajectories and compute confidence bounds, and a high-rate optimization-based controller that runs at higher frequency to track these trajectories with smooth adjustments. This segmentation allows high responsiveness where needed while reducing overall computational load.
Solution Approach 2:
The patent applies different levels of computational effort to different aspects of control. The high-rate controller focuses computational resources on local trajectory tracking and smoothness adjustments where responsiveness is critical, while relying on the pre-computed confidence bounds from the low-rate planner for global path guidance, thus achieving high responsiveness without excessive overall computational load.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A control system controls a vehicle using a probabilistic motion planner and an adaptive predictive controller. The probabilistic motion planner produces a sequence of parametric probability distributions over a sequence of target states for the vehicle with parameters defining a first and higher order moments. The adaptive predictive controller optimizes a cost function over a prediction horizon to produce a sequence of control commands to one or multiple actuators of the vehicle. The cost function balances a cost of tracking of different state variables in the sequence of the target states defined by the first moments. The balancing is performed by weighting different state variables using one or multiple of the higher order moments of the probability distribution.