Automated Parking Path Planning With RL Segment Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle path planning algorithms require significant computational resources and energy, struggle with sharp maneuvers that cause passenger discomfort, and are challenging to generate acceptable paths within limited on-board processing power and time constraints.
Innovation Solution
A method using reinforcement learning to iteratively generate vehicle path segments, combining simple and complex actions, and employing a neural network to estimate path scores, allowing for efficient path planning within computational limits and real-time adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional path planning algorithms are used, then path generation accuracy is improved, but computational load and energy consumption increase significantly
Solution Approach 1:
The patent segments the continuous path planning problem into discrete path segments that can be pre-computed and stored. By dividing the complex path generation task into manageable segments, the system achieves accurate path planning while reducing real-time computational energy requirements, as segments can be selected and combined from pre-computed options rather than generating entire paths from scratch.
Solution Approach 2:
The patent performs preliminary computation of path segments offline or in advance, storing them for later use. This preliminary action allows the system to have accurate path options pre-prepared, eliminating the need for heavy real-time computation when actual path planning is needed, thus reducing energy consumption during vehicle operation while maintaining path accuracy.
2Measurement precision
If traditional path planning algorithms are used, then path generation accuracy is improved, but processing time increases
Solution Approach 1:
By segmenting the path into pre-computed components, the system avoids time-consuming real-time generation of entire paths. The segmentation allows rapid assembly of accurate path solutions from stored segments, significantly reducing processing time while maintaining the accuracy benefits of traditional algorithms.
Solution Approach 2:
The preliminary computation and storage of path segments enables the system to have accurate path options ready in advance. When path planning is needed, the system can quickly retrieve and combine pre-computed segments rather than performing lengthy real-time calculations, thus achieving both accuracy and speed.
3Measurement precision
If path planning includes sharp turns and frequent stops to reach desired position, then positioning accuracy is improved, but passenger comfort deteriorates
Solution Approach 1:
The patent applies parameter changes by transforming the path representation into a continuous function with adjustable parameters for smoothness, velocity, and acceleration. This allows the system to optimize paths for both positioning accuracy and passenger comfort by adjusting these parameters to eliminate sharp turns and jerky movements while still achieving the desired final position.
Solution Approach 2:
The patent introduces dynamics by using continuous path functions that can be smoothly adjusted and optimized. Rather than relying on discrete, rigid path segments with sharp transitions, the dynamic continuous functions allow for smooth variations in direction and speed, ensuring passenger comfort while maintaining positioning accuracy through parameter optimization.
4Measurement precision
If more computational resources are allocated to path planning, then path quality is improved, but vehicle controller complexity increases
Solution Approach 1:
The patent reduces controller complexity by segmenting the path planning task into pre-computed segments stored in memory. The controller only needs to select and combine these segments based on current conditions, rather than implementing complex real-time path generation algorithms, thus maintaining high path quality with simpler controller architecture.
Solution Approach 2:
By performing complex path computation in advance and storing results, the patent eliminates the need for sophisticated real-time algorithms in the vehicle controller. The preliminary action of pre-computing paths shifts the computational burden away from the controller, allowing it to maintain high path quality through simple selection and combination of pre-prepared segments.
Data Source
AI summary
A method for planning a vehicle path includes generating an initial perception data set using one or more vehicle sensors. The data set includes data defining spatial positions of features extrinsic to the vehicle. The method determines an operational area of the parking operation, at least one goal pose of the vehicle, and a set of constraints. The method iteratively generates path segments using a reinforcement learning algorithm. Each completed set of path segments is configured to reposition the vehicle from the initial pose to one of the goal poses. Iteratively generating the sets of path segments includes determining a total path score for each generated set of path segments. The method selects a set of path that has a best total path score, generates a set of path points and provides the set of path points to an automated parking controller operation.