Vehicle Collision Avoidance Using Lidar-Based Object Contours
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle navigation systems often rely on bounding boxes for object representation, which can be inaccurate, leading to potential collisions due to their inability to accurately depict the shape and size of objects, especially those with irregular shapes or protrusions.
Innovation Solution
The use of bounding contours generated from lidar data, which project lidar point clouds onto a two-dimensional plane and create a convex hull to determine a tight-fitting perimeter of objects, allowing for more accurate collision detection and trajectory planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a bounding box is used to represent an object, then the object representation is simple and computationally efficient, but the accuracy of object shape and size representation deteriorates
Solution Approach 1:
The patent segments the object representation into two parts: a bounding box for coarse collision detection and a bounding contour for precise shape representation. This segmentation allows the system to use the simple bounding box for initial fast checks while employing the more accurate bounding contour only when needed for final collision determination, thus resolving the contradiction between computational efficiency and measurement precision.
Solution Approach 2:
The patent transitions from representing objects in one dimension (bounding box with width and height) to two dimensions (bounding contour with actual object perimeter). This dimensional enhancement allows the system to capture irregular shapes and protrusions that cannot be represented by a rectangular bounding box, thereby improving measurement precision while maintaining computational efficiency through selective usage.
2Speed
If a bounding box is used for collision detection, then the detection process is fast, but collision detection accuracy deteriorates due to inability to represent irregular shapes
Solution Approach 1:
The patent divides the collision detection process into two stages: a fast bounding box check for initial filtering and a more accurate bounding contour check for final determination. This segmentation enables the system to maintain high detection speed by using the simple bounding box for most objects while ensuring high accuracy by applying the bounding contour when the bounding box indicates potential collision or when the object has irregular shapes.
Solution Approach 2:
The patent performs a preliminary collision check using the bounding box before conducting the final check with the bounding contour. This preliminary action allows the system to quickly eliminate objects that cannot possibly collide with the vehicle, thereby maintaining fast detection speed while ensuring accurate collision detection for objects that require detailed shape analysis.
3Measurement precision
If a tight-fitting bounding contour is generated from lidar data, then the object representation accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent segments the computational workload by generating bounding contours only for objects that require high-precision representation, such as objects with irregular shapes or those detected as potential collisions by the bounding box. For other objects, the system uses the simpler bounding box representation, thereby reducing overall computational complexity while maintaining high measurement precision where needed.
Solution Approach 2:
The patent applies the computationally intensive bounding contour generation partially, only to objects where it provides necessary value. Instead of generating bounding contours for all detected objects, the system selectively applies this technique to objects that benefit most from accurate shape representation, thus balancing measurement precision improvement with computational complexity management.
Data Source
AI summary
Techniques for collision avoidance using an object contour are discussed. A trajectory associated with a vehicle may be received. Sensor data can be received from a sensor associated with the vehicle. A bounding contour may be determined and associated with an object represented in the sensor data. Based on the trajectory, a simulated position of the vehicle can be determined. Additionally, a predicted position of the bounding contour can be determined. Based on the simulated position of the vehicle and the predicted position of the bounding contour, a distance between the vehicle and the object may be determined. An action can be performed based on the distance between the vehicle and the object.