Autonomous Vehicle Trajectory Control Using Space-Time Corridors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for controlling autonomous or semi-autonomous vehicles face challenges in efficiently optimizing motion trajectories while considering time-dependent drivable spaces and obstacles, often leading to complex computations and suboptimal results.
Innovation Solution
A method involving a two-part optimization approach, first defining a space-time corridor using convex objects to approximate the drivable space, followed by a nonlinear programming solver to optimize the motion trajectory, utilizing a blended vehicle model for different speed ranges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a two-part optimization approach with space-time corridor and convex objects is used, then computational complexity is reduced and globally optimal trajectories are achieved, but the device complexity and algorithm complexity increase
Solution Approach 1:
The optimization process is divided into two distinct parts: first defining a space-time corridor using convex objects to approximate drivable space, then applying a nonlinear programming solver to optimize the motion trajectory. This segmentation allows each part to focus on specific aspects of the problem, reducing overall computational complexity while maintaining solution quality.
Solution Approach 2:
The patent introduces a space-time corridor that adds the time dimension to the traditional spatial representation of drivable space. By transforming the problem into a four-dimensional space-time framework and using convex objects to approximate this space, the method enables more efficient optimization while capturing temporal constraints and dynamics.
2Ease of operation
If a blended vehicle model is used for different speed ranges, then driving performance is improved and abrupt changes are reduced, but the model complexity increases
Solution Approach 1:
The patent employs a blended vehicle model that uses different model representations for different speed ranges. By selecting appropriate model characteristics locally based on the operating conditions (speed range), the system achieves improved driving performance and reduced abrupt changes without requiring a single overly complex model to handle all scenarios.
Solution Approach 2:
The vehicle model is made dynamic by blending different model representations based on the current operating conditions, particularly speed ranges. This allows the model to adapt its characteristics in real-time, providing more accurate and appropriate behavior for different driving scenarios while maintaining computational efficiency.
3Productivity
If convex objects are used to approximate time-dependent drivable space, then computational efficiency is improved and safety is enhanced, but the precision of space representation may be reduced
Solution Approach 1:
The patent uses convex objects to approximate the time-dependent drivable space, creating a slightly conservative representation that ensures safety by beforehand cushioning against potential precision errors. This approximation provides a computationally efficient framework that maintains safety margins while enabling real-time optimization.
Data Source
AI summary
The present invention is directed to computer-implemented method for controlling an autonomous or semi-autonomous vehicle. A vehicle model is provided and information describing a roadway (301) is obtained. This information is transformed in a Frenet-Serret frame, which defines a curvilinear coordinate space. Then, in the curvilinear coordinate space, a plurality of line segments (306, 307) are defined (3) along the roadway (301), at locations spaced longitudinally along a reference line defining the Frenet-Serret frame. The line segments have respective lengths that extends, each, perpendicularly to the reference line. Next, based on the defined line segments, a space-time corridor (327, 327″, 326″) is determined (5) to obtain a time-dependent space available for the vehicle. Bounds for the vehicle within said time-dependent space are subsequently obtained (6). The bounds define a set of constraints that delimit a plurality of convex space-time objects along the reference line. Said objects approximate the time-dependent space. Then, based on the obtained set of constraints and the provided vehicle model, the method provides (7) a motion trajectory of the vehicle through the plurality of convex space-time objects. This motion trajectory is optimized using a nonlinear programming solver. Finally, the motion trajectory is transformed (8) from the curvilinear coordinate space to a Euclidean coordinate space, to control the vehicle based on the transformed motion trajectory. The present invention further concerns related systems and computer program products.