Radar Velocity Estimation for Low-Latency Vehicle Perception
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional navigation systems for autonomous vehicles rely heavily on image data, which can lead to increased processing times and latency, particularly when detecting and predicting the motion of objects that change trajectory quickly.
Innovation Solution
A radar-based perception system that processes radar data in conjunction with or instead of image data, utilizing machine learning models to convert raw radar point cloud data into object-level representations, thereby reducing latency and improving processing speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional perception systems use image data for object detection and motion prediction, then object recognition capability is improved, but processing time increases and latency is worsened
Solution Approach 1:
The patent replaces image-based perception with radar-based perception. Radar sensors emit electromagnetic waves and detect reflections from objects, providing direct measurements of position, velocity, and acceleration without requiring complex image processing. This substitution of the sensing mechanism fundamentally reduces processing time while maintaining detection accuracy, as radar data can be processed more quickly than high-resolution images.
Solution Approach 2:
The patent changes the fundamental parameters being measured from optical intensity (image data) to electromagnetic wave reflection characteristics (radar data). Radar directly measures range, range rate (velocity), and acceleration through phase and frequency analysis, bypassing the need for complex image processing algorithms. This parameter change enables faster processing while providing motion information that is more directly useful for prediction.
2Reliability
If conventional perception systems rely on image data, then object detection is improved, but system latency is worsened
Solution Approach 1:
The system substitutes radar sensing for image-based sensing. Radar provides direct geometric and kinematic measurements through electromagnetic wave propagation, eliminating the latency associated with image capture, preprocessing, feature extraction, and object detection algorithms. This substitution maintains detection reliability while significantly reducing system latency.
Solution Approach 2:
The patent performs preliminary registration of radar data to a global reference frame, which establishes accurate spatial and temporal relationships before further processing. This preliminary action of aligning data in the global frame enables more efficient subsequent processing and reduces overall system latency by preventing rework and iterations.
3Productivity
If radar data is used for perception, then processing speed is improved and latency is reduced, but performance in all environmental conditions is not guaranteed
Solution Approach 1:
The patent employs multiple radar sensors positioned at different locations on the autonomous vehicle, each providing a different viewing angle and coverage area. This multi-functional sensor array ensures that objects remain detectable even when obscured from one viewpoint or when environmental conditions affect specific radar beams. The system processes data from all sensors to maintain consistent performance across diverse environmental conditions.
Solution Approach 2:
The system uses feedback from multiple radar sensors and iterative data association algorithms to maintain reliable object tracking. By continuously comparing measurements from successive radar scans and adjusting data associations accordingly, the system compensates for environmental variations and maintains consistent detection and tracking performance across different weather and lighting conditions.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The radar-based perception system enhances the operational safety of autonomous vehicles by reducing latency, improving performance in low-light environments, and providing accurate range rate measurements, especially in degraded conditions such as fog, rain, or snow.
Implementation Method 1
a radar-based perception system which may be implemented using radar data in lieu of or in addition to other types of sensor data
Implementation Method 2
The radar being phase coherent... produces improved accuracy with respect to determining the range rate or relative velocity of objects detected in the scene
Data Source
AI summary
Techniques for updating data operations in a perception system are discussed herein. A vehicle may use a perception system to capture data about an environment proximate to the vehicle. The perception system may output the data about the environment to a system configured to determine positions of objects relative to the perception system over time. The positions of the objects may be used to estimate an object velocity and may be compared against machine learning model outputs in a self-supervised manner to train the machine learning model to output object velocities based on inputs from the perception system. The output of the machine learning model may include a two-dimensional velocity for objects in the environment. The two-dimensional velocity may be used for a vehicle system such that the vehicle can make environmentally aware operational decisions, which may improve reaction time(s) and/or safety outcomes of the vehicle.