Attention Masked Motion Planning for Autonomous Vehicle Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in efficiently utilizing processing resources for motion planning by focusing on irrelevant areas of sensor data, leading to increased computational complexity and reduced efficiency in object detection and trajectory selection.
Innovation Solution
The use of a binary attention mask generated by a machine-learned model to prioritize important areas in sensor data, allowing for efficient resource allocation and improved motion planning by focusing on relevant data through a multi-stage machine-learned model trained with a joint multi-task learning objective.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the autonomous vehicle processes all sensor data without filtering, then comprehensive environmental awareness is achieved, but computational complexity increases and processing efficiency decreases
Solution Approach 1:
The patent segments the sensor data processing by dividing the point cloud into multiple subsets based on spatial regions. Each subset is processed independently by separate neural network models, allowing parallel computation and reducing the computational burden on a single system while maintaining comprehensive environmental awareness.
Solution Approach 2:
The patent extracts and processes only the relevant portions of sensor data by generating attention masks that identify important regions. This extraction principle allows the system to focus computational resources on critical areas of the environment while ignoring irrelevant data, thereby improving efficiency without sacrificing safety.
2Measurement precision
If the autonomous vehicle uses a comprehensive machine-learned model to analyze all areas, then object detection accuracy is improved, but energy consumption increases
Solution Approach 1:
The patent applies partial action by processing only a subset of sensor data that is most relevant to the current driving context. Through attention mechanisms and spatial region segmentation, the system performs sufficient processing on critical areas while reducing or skipping processing in less important regions, thereby maintaining detection accuracy while lowering energy consumption.
3Speed
If the autonomous vehicle focuses processing on specific areas using attention masks, then processing speed is improved, but coverage of monitored areas may be reduced
Solution Approach 1:
The patent divides the monitoring area into multiple spatial segments or regions, each processed by dedicated computational units. This segmentation enables parallel processing across different areas, maintaining both high processing speed and comprehensive coverage by ensuring that no single bottleneck limits the overall system performance.
Solution Approach 2:
The patent creates a multi-functional processing system where the same neural network architecture and attention mechanisms are applied across multiple spatial regions. This universal approach allows the system to handle diverse areas (foreground, background, different distance zones) with consistent processing quality while maintaining efficiency through standardized reusable components.
Data Source
AI summary
Systems and methods for generating attention masks are provided. In particular, a computing system can access sensor data and map data for an area around an autonomous vehicle. The computing system can generate a voxel grid representation of the sensor data and map data. The computing system can generate an attention mask based on the voxel grid representation. The computing system can generate, by using the voxel grid representation and the attention mask as input to a machine-learned model, an attention weighted feature map. The computing system can determine using the attention weighted feature map, a planning cost volume for an area around the autonomous vehicle. The computing system can select a trajectory for the autonomous vehicle based, at least in part, on the planning cost volume.