Autonomous Vehicle Trajectory Optimization via Dynamic Programming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving vehicles face challenges in optimizing motion planning and control due to open-ended operations without initial constraints, leading to inaccurate and rough navigation.
Innovation Solution
The implementation of a decision and planning system that uses dynamic programming and gradient descent optimization to calculate optimal trajectories for autonomous vehicles, considering obstacles and traffic rules, by generating path and speed profiles and applying quadratic programming and gradient descent methods to minimize costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If motion planning and control are applied without initial constraints, then the system maintains high adaptability to different driving scenarios, but the navigation accuracy and smoothness deteriorate
Solution Approach 1:
The patent applies preliminary action by establishing initial constraints on motion planning before the optimization process begins. These constraints include predefined safety margins, road boundary limitations, and traffic rule compliance requirements that are set in advance to guide the optimization algorithm, thereby ensuring navigation accuracy while maintaining adaptability to different driving scenarios.
Solution Approach 2:
The patent utilizes parameter changes by dynamically adjusting optimization parameters such as cost function weights, constraint thresholds, and penalty factors based on the specific driving scenario. This allows the system to adapt to different situations (urban, highway, rural) while maintaining accurate navigation through controlled parameter modifications rather than changing the fundamental planning approach.
2Adaptability or versatility
If general motion planning algorithms are used for all vehicle types, then the system achieves broad applicability, but the navigation smoothness and accuracy deteriorate under specific circumstances
Solution Approach 1:
The patent applies local quality by implementing vehicle-specific parameters and constraints within the general motion planning framework. Different vehicle types (autonomous vehicles, trailers, combinations) have customized parameters such as wheelbase length, maximum steering angle, and acceleration limits that are locally adjusted to match their specific characteristics, thereby achieving both broad applicability and vehicle-specific navigation smoothness.
Solution Approach 2:
The patent utilizes dynamics by making the motion planning algorithm adaptive to different vehicle configurations through dynamic parameter adjustment. The system can switch between different vehicle models and their associated parameters based on the current vehicle type, allowing the same base algorithm to achieve optimal navigation smoothness for each specific vehicle configuration.
3Manufacturing precision
If dynamic programming and gradient descent optimization are applied, then the trajectory accuracy and collision avoidance improve, but the computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the motion planning problem into multiple discrete stages or time steps, where dynamic programming can be efficiently applied. The continuous trajectory optimization is broken down into segmented decisions at each time step, allowing the use of efficient DP algorithms while maintaining high trajectory accuracy through fine-grained temporal resolution.
Solution Approach 2:
The patent replaces traditional mechanical or rule-based motion planning approaches with gradient descent optimization methods. This substitution allows for more accurate and smooth trajectory generation by using continuous mathematical optimization instead of discrete rule-based decisions, achieving higher trajectory accuracy while the computational complexity is managed through efficient gradient-based algorithms.
Data Source
AI summary
According to some embodiments, a system calculates a first trajectory based on a map and a route information. The system generates a path profile based on the first trajectory, traffic rules, and an obstacle information describing one or more obstacles perceived by the ADV, where for each of the obstacles, the path profile includes a decision to yield or nudge to left or right of the obstacle. The system generates a speed profile based on the path profile in view of the traffic rules. The system performs a gradient descent optimization based on the path profile and the speed profile to generate a second trajectory representing an optimized first trajectory and controls the ADV according to the second trajectory.