Predictive Occupancy Grid Filtering for Autonomous Vehicle Path Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing path planning methods for autonomous vehicles face challenges in representing dynamic objects in real-world environments, leading to inefficient navigation and frequent trajectory modifications in dynamic, unstructured environments.
Innovation Solution
A processor-implemented method generates a single predicted 2D occupancy grid map by filtering a set of predicted occupancy grid maps using weight maps and area of interest maps, based on kinodynamic parameters and behavior commands, to reduce computational complexity and improve trajectory planning in dynamic environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single occupancy grid map is used for path planning, then the path planning computation is simple and fast, but the planning is only valid for static environments and fails in dynamic environments
Solution Approach 1:
The patent transforms the traditional 2D occupancy grid map into a 3D spatiotemporal occupancy grid map by adding the time dimension. This allows the system to represent not only spatial occupancy but also temporal evolution of the environment, enabling path planning that accounts for moving objects while maintaining computational efficiency through a unified 3D representation.
Solution Approach 2:
The system performs preliminary actions by predicting future positions of moving objects and pre-computing the 3D spatiotemporal occupancy grid map before actual path planning. This predictive approach allows the planner to account for dynamic object movements in advance, reducing the need for frequent trajectory modifications during execution.
2Adaptability or versatility
If multiple occupancy grid maps are generated for different time steps to account for moving objects, then the environment representation becomes more accurate for dynamic environments, but the computational complexity and processing time increase significantly
Solution Approach 1:
The patent merges multiple 2D occupancy grid maps across different time steps into a single 3D spatiotemporal occupancy grid map. This consolidation integrates spatial and temporal information into one unified data structure, allowing the system to handle dynamic environments without the computational burden of processing multiple separate maps independently.
Solution Approach 2:
By transitioning from multiple 2D maps to a single 3D spatiotemporal map, the system efficiently packs temporal information into the third dimension. This dimensional transformation reduces computational complexity while preserving all necessary information about moving objects across time.
3Reliability
If the occupancy grid map is updated frequently with new sensor data, then the environment information remains current and accurate, but the planned trajectory requires frequent modifications and unnecessary steering corrections
Solution Approach 1:
The system performs preliminary prediction of moving object trajectories and pre-updates the 3D spatiotemporal occupancy grid map to reflect future environmental states. This allows the path planner to work with a stable, forward-looking representation that reduces the need for frequent trajectory re-planning and steering corrections during execution.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A processor-implemented method and system for determining a predictive occupancy grid map (OGM) for an autonomous vehicle are disclosed. The method includes: receiving a set of OGMs including a current predicted OGM and one or more future predicted OGMs, the current OGM associated with a current timestamp and each future predicted OGM associated with a future timestamp; generating a weight map associated with the current timestamp based on one or more kinodynamic parameters of the vehicle at the current time stamp, and one or more weight map associated with a future timestamp; generating a set of filtered predicted OGMs by filtering the current predicted OGM with the weight map associated the current timestamp and filtering each respective future predicted OGM associated with a future timestamp with the weight map associated with the respective future timestamp; and sending a single predicted OGM to a trajectory generator.