In-path Object Detection Using Inertial and Radar Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing safety systems in autonomous and semi-autonomous vehicles face challenges in accurately determining the lane navigation and object positioning due to incomplete or absent lane signatures in image data, requiring extensive computing resources and multiple types of sensors.
Innovation Solution
The system uses inertial measurement unit data to determine the vehicle's motion and lane curvature, generating lane boundaries without relying on image data, and employs RADAR and LiDAR data to assess object locations relative to the vehicle's path, reducing the need for image sensors and processing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image data processing is performed to determine lane signatures and object locations, then object detection accuracy is improved, but computing resource consumption increases
Solution Approach 1:
The patent extracts and removes the image processing component from the safety system, relying instead on pre-determined lane information from map data and object location data from sensors like RADAR and LiDAR. This extraction eliminates the computational burden of image processing while maintaining the ability to determine whether objects are in the vehicle's lane.
Solution Approach 2:
The patent substitutes the mechanical/image processing system with a data-based approach using map data and sensor data. Instead of processing images to identify lane signatures, the system uses pre-stored lane geometry data combined with sensor object locations to determine in-lane objects, replacing computationally intensive image processing with simpler data comparison operations.
2Reliability
If multiple types of sensors (image sensors, RADAR, LiDAR) are used to detect lanes and objects, then detection reliability is improved, but device complexity increases
Solution Approach 1:
The patent extracts and removes the image sensor requirement from the system. By using pre-determined lane information from map data combined with object location data from RADAR or LiDAR, the system achieves reliable lane and object detection without needing image sensors, thereby reducing the number and types of sensors required.
Solution Approach 2:
The patent uses map data as a copy or representation of lane information, eliminating the need for real-time image capture and processing. The map data provides a digital replica of lane geometry that can be combined with sensor data to determine object locations relative to the lane, reducing sensor complexity while maintaining detection reliability.
3Measurement precision
If extensive image processing and sensor data processing are performed, then lane detection accuracy is improved, but runtime increases
Solution Approach 1:
The patent applies preliminary action by pre-determining and storing lane information in map data before the vehicle reaches those locations. This allows the system to skip real-time image processing and directly compare sensor-detected object locations with pre-stored lane geometry, significantly reducing runtime while maintaining lane detection accuracy.
Solution Approach 2:
The patent substitutes real-time image processing with a data comparison approach using pre-stored map data. Instead of processing images to identify lane signatures during runtime, the system compares sensor object locations with pre-determined lane boundaries from map data, eliminating computationally intensive operations and reducing system runtime.
Data Source
AI summary
In various examples, in-path object detection for autonomous and semi-autonomous systems and applications is described herein. For instance, systems and methods may use sensor data, such as inertial measurement sensor data, to generate or determine a lane that is associated with a vehicle. In some examples, the lane is generated or determined by determining a curvature of a path that the vehicle is navigating and then adding distance boundaries to both sides of the path. The systems and methods may then use sensor data, such as RADAR data and/or LiDAR data, to determine one or more locations of one or more objects with respect to the vehicle. Using the geometry of the lane and the location(s) of the object(s), the systems and methods may then determine whether the object(s) is located along the path of the vehicle.