Vehicle Local Path Planning Around Parked and Oncoming Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in adapting their local path planning to avoid parked vehicles on the roadside, leading to prolonged waiting times and potential collisions with oncoming traffic.
Innovation Solution
A local path planning system that classifies objects as upper or lower bounds based on bounding box distances and cost functions, generating trajectories using a Frenet frame and non-linear kinematic bicycle model, incorporating curvature-based corrections and bounded slack variables to handle perception noise and dynamic obstacles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the autonomous vehicle waits for parked vehicles to move before continuing on a pre-determined path, then collision avoidance with parked vehicles is achieved, but the vehicle experiences extended waiting time and reduced productivity
Solution Approach 1:
The system performs preliminary classification of objects as upper bound, lower bound, or infeasible based on bounding box distances and cost functions before path planning. This pre-processing of environmental data allows the vehicle to proactively identify navigable paths around parked vehicles rather than passively waiting, resolving the contradiction by preparing collision avoidance information in advance while maintaining continuous motion
Solution Approach 2:
The path planning system dynamically adapts the reference path in real-time based on classified objects and generated boundaries. Instead of following a static pre-determined path that requires waiting, the system continuously updates the trajectory to deviate around parked vehicles and return to the reference path, enabling dynamic collision avoidance without extended waiting time
2Reliability
If the autonomous vehicle deviates from the lane center to avoid parked vehicles, then collision avoidance is achieved, but the vehicle may collide with oncoming traffic
Solution Approach 1:
The system incorporates feedback mechanisms by continuously monitoring the operating environment, classifying new objects, and updating boundaries in real-time. The path planning uses this feedback to adjust the trajectory dynamically, ensuring that deviations from the lane center are controlled and reversible, allowing the vehicle to avoid parked vehicles while maintaining awareness of oncoming traffic and correcting the path when safe
Solution Approach 2:
The system changes path planning parameters by transforming the kinematic model from space-time domain to space-only domain and using Frenet frame coordinates. This parameter transformation enables more precise control over lateral deviation from the reference path, allowing the vehicle to navigate around parked vehicles with controlled excursions that minimize the risk of collision with oncoming traffic
3Measurement precision
If the system uses a detailed kinematic model for accurate path planning, then navigation precision is improved, but computational complexity increases
Solution Approach 1:
The system segments the path planning problem into distinct components: object classification based on bounding box distances, boundary generation from classified objects, and trajectory optimization using the segmented reference path. This segmentation allows each component to be processed independently with appropriate computational methods, maintaining high accuracy while reducing overall computational complexity through modular processing
Solution Approach 2:
The system performs parameter changes by transforming the kinematic model from space-time domain to space-only domain and using Frenet frame coordinates. This transformation simplifies the mathematical representation of the path planning problem, reducing computational complexity while preserving the accuracy needed for precise navigation around dynamic obstacles
Data Source
AI summary
According to one aspect, local path planning may include classifying an object within an operating environment as an upper bound object, a lower bound object, or infeasible based on a distance between a bounding box associated with the object and an upper environment feature, a distance between the bounding box associated with the object and a lower environment feature, and a cost function, generating a boundary associated with the object and the upper environment feature or the lower environment feature based on the classification of the object and boundary points of the bounding box associated with the object, and generating a local path planning trajectory for a vehicle based on the classification of the object, the boundary associated with the object, and transforming a kinematic model from a space-time domain to a space-only domain.