Autonomous Parking Trajectory Planning With Obstacle Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicle parking in dynamic environments, such as parking areas with obstacles, is challenging due to the need for accurate short-term and long-term motion prediction of obstacle vehicles, and existing methods lack consistency and fail to integrate short-horizon and long-horizon planning effectively, especially in the presence of rapidly changing obstacles.
Innovation Solution
A system and method that includes an environment predictor to predict both the path and mode of motion of obstacle vehicles, determining safety constraints based on these predictions, and a strategic motion planner that adapts control strategies using model predictive control, search-based retreat planning, and optimization-based path repair to ensure collision-free parking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data driven methods (IMM or classifiers) are used for motion prediction, then short-term trajectory prediction can be achieved, but prediction accuracy is inconsistent and error increases with poor data set selection
Solution Approach 1:
The motion prediction problem is segmented into two distinct components: short-term motion prediction (using physics-based dynamic models for immediate trajectory) and long-term mode prediction (using classifiers for behavioral patterns). This segmentation allows each component to be optimized independently, with the short-term predictor providing consistent accuracy through physics-based modeling rather than data-driven approaches.
Solution Approach 2:
A hierarchical prediction framework is introduced as an intermediary between raw sensor data and motion planning. The framework uses long-term mode prediction to identify behavioral patterns first, then feeds this information to the short-term motion prediction module. This intermediary structure improves overall prediction consistency by combining complementary approaches rather than relying on a single data-driven method.
2Adaptability or versatility
If general motion planning algorithms are used, then they can handle various scenarios, but they fail to integrate short-horizon planning with long-horizon planning effectively for parking
Solution Approach 1:
The motion planning process is segmented into long-horizon planning (strategic path generation considering parking goals and obstacle patterns) and short-horizon planning (tactical trajectory refinement for immediate maneuvers). This segmentation enables effective integration of different time scales, with each planner optimized for its specific temporal scope while maintaining overall system versatility.
Solution Approach 2:
The planning system dynamically switches between different planning horizons based on the situation. When obstacles are detected or the environment changes, the system transitions to short-horizon planning for rapid reaction. When the environment is stable, it uses long-horizon planning for efficient path generation. This dynamic adaptation maintains both versatility and efficiency.
3Ease of operation
If motion planners specialized in autonomous parking are used, then parking maneuver capability is improved, but they cannot adequately incorporate online path repairing upon new obstacles
Solution Approach 1:
The system implements continuous feedback loops where sensor data about new obstacles is constantly monitored during parking maneuvers. When obstacles are detected, the feedback triggers online path repair processes that adjust the trajectory in real-time. This feedback mechanism enables specialized parking planners to adapt to new situations while maintaining their maneuver capability.
Solution Approach 2:
The planning system dynamically adapts its behavior based on real-time obstacle detection. When new obstacles appear, the system transitions from executing pre-planned trajectories to performing online path repair. This dynamic response allows the specialized parking planner to maintain its maneuver capability while gaining adaptability to unexpected situations.
Data Source
AI summary
The present disclosure provides a system and a method for parking an autonomous ego-vehicle in a dynamic environment of a parking area. The method includes collecting measurements of a state of the dynamic environment, a state of one or multiple stationary vehicles and one or multiple obstacle vehicles moving in the parking area. The method further includes executing a path planner configured to produce a trajectory based on the state of the dynamic environment and executing an environment predictor configured to predict a path and a mode of motion for each of the obstacle vehicles. The method further includes determining a safety constraint for each of the obstacle vehicles based on the path and the mode of motion for each of the obstacle vehicles and parking the autonomous ego-vehicle based on the trajectory for parking and the safety constraint for each of the obstacle vehicles.