Route-Relative Trajectory Integration for Singularity-Free AV Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining and tracking trajectories for autonomous vehicles are inefficient and prone to instability, especially in resource-limited or time-limited environments, due to high curvature roadways and singularities, which can lead to inaccurate control and safety issues.
Innovation Solution
A numerical integrator is employed for a second-order kinematic vehicle model in a route-relative frame, allowing for the prediction of vehicle states and the determination of stable trajectories by avoiding singularities and using a backstepping controller for initialization, thus ensuring accurate and efficient trajectory planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional trajectory methods are used for autonomous vehicles, then the vehicle can follow a route, but the system becomes unstable and inaccurate when encountering high curvature roadways or perpendicular orientations due to singularities
Solution Approach 1:
The patent transforms the traditional body-frame integration approach into a route-frame integration approach. Instead of integrating vehicle dynamics in the vehicle's local coordinate system (which causes singularities at perpendicular orientations), the system integrates in the route's coordinate system where the route pose serves as the independent variable. This inversion of the reference frame eliminates the singularity problem because the route frame naturally accommodates high curvature roadways without encountering the perpendicular orientation issue that plagues body-frame methods.
Solution Approach 2:
The patent changes the fundamental parameters of the integration system by redefining the independent variable from body-frame position to route pose (arc length along the route). By parameterizing the vehicle state relative to the route progression rather than absolute position, the system can handle high curvature roadways smoothly. The route-curvature parameter is introduced to explicitly model the road geometry, allowing the integrator to adapt to varying curvature without encountering singularities.
2Manufacturing precision
If separate optimization of vehicle acceleration and steering angle is performed, then the trajectory can be planned, but the computational efficiency decreases in resource-limited or time-limited environments
Solution Approach 1:
The patent merges the separate optimization processes for acceleration and steering angle into a unified route-relative trajectory integrator. Instead of independently optimizing acceleration and steering as separate control inputs, the system formulates a single integration process that simultaneously determines both parameters based on the desired route progression. This unified approach reduces the computational burden by eliminating redundant calculations and iterations, while maintaining trajectory planning accuracy through the coupled determination of motion and orientation.
Solution Approach 2:
The route-relative integrator serves multiple functions simultaneously: it integrates vehicle dynamics, plans the trajectory, and adapts to varying road curvature all within a single computational framework. By making the integrator route-relative, it universally handles different road geometries (straight roads, curved roads, high curvature roadways) without requiring separate processing logic, thereby improving computational efficiency across diverse driving scenarios while maintaining accurate trajectory planning.
3Productivity
If route-relative numerical integration is implemented, then computational load is reduced and stability is improved, but the implementation complexity increases due to the need for route pose parameterization
Solution Approach 1:
The patent applies preliminary action by pre-parameterizing the route into discrete pose elements before the integration process begins. The route is预先 divided into segments with defined pose characteristics (position, orientation, curvature) that serve as the foundation for the numerical integration. This preprocessing step simplifies the actual integration computation by providing ready-to-use route parameters, reducing the computational complexity during real-time execution despite the initial setup effort.
Data Source
AI summary
Generating a lane reference from a roadway shape and/or generating a trajectory for controlling an autonomous vehicle may include determining a predicted state of the lane reference and/or a candidate trajectory by an integrator. The disclosed integrator is implemented as a numerical integrator in predominantly closed-form that is able to avoid singularities while maintaining no approximation error. The disclosed integrator is also more robust to poor initial estimations, high curvature roadways, and zero-velocity conditions.