Autonomous Vehicle Object Detection Beyond LiDAR Range
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in detecting and responding to objects beyond the range of their LIDAR sensors, which can result in insufficient time to change lanes, especially on highways, and are hindered by occlusions from larger objects like tractor-trailers.
Innovation Solution
The use of high-resolution cameras for initial object detection and tracking, with machine learning models to identify and compare bounding boxes across images, and fusion with data from LIDAR and radar sensors as objects come closer, allowing for long-range object detection and timely lane changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If LIDAR sensors are used for object detection, then detection reliability is improved within 100 meters, but detection range is limited and objects beyond this range cannot be detected in time
Solution Approach 1:
The patent combines multiple sensor types (LIDAR, radar, cameras) with different detection ranges and capabilities into a unified sensor fusion system. LIDAR provides reliable short-range detection, radar extends medium-range detection, and cameras provide long-range visual information, creating a comprehensive detection system that overcomes the limited range of individual sensors.
Solution Approach 2:
The detection space is segmented into different ranges with different sensor types optimized for each range. Short-range detection (0-100m) uses LIDAR for high reliability, medium-range (100-300m) uses radar, and long-range (>300m) uses cameras, allowing each sensor to operate in its optimal performance zone.
2Length of stationary object
If high-resolution cameras are used for long-range detection, then detection range is extended to 300 meters, but detection precision and reliability decrease compared to LIDAR
Solution Approach 1:
Cameras perform preliminary long-range detection to identify potential objects at distances beyond LIDAR range. When objects are detected by cameras at long range, the system proactively begins tracking and monitoring these objects, preparing for potential hazards before they enter the high-precision LIDAR detection zone.
Solution Approach 2:
Radar serves as an intermediary sensor that bridges the gap between camera-based long-range detection and LIDAR-based short-range precision detection. Radar provides medium-range detection with better precision than cameras but longer range than LIDAR, creating a seamless transition zone in the detection spectrum.
3Ease of operation
If forward-facing LIDAR is used on large vehicles like class 8 trucks, then detection capability is provided, but LIDAR becomes occluded by larger objects such as other tractor trailers
Solution Approach 1:
The system employs multiple sensor types that can detect objects in different configurations and positions. Cameras mounted at various locations (front, rear, sides) and radar sensors provide detection capabilities that are less susceptible to occlusion, ensuring that at least some sensors can detect objects regardless of blocking by other large vehicles.
Solution Approach 2:
The system transitions from relying solely on forward-facing sensors to a multi-dimensional sensor arrangement including side-mounted and rear-mounted sensors. This spatial distribution across multiple dimensions allows detection of objects that may be occluded from the forward view, such as vehicles stopped in adjacent lanes or merging from side roads.
4Reliability
If lane change maneuver is performed to avoid stopped vehicle on shoulder, then safety is improved, but maneuver time must be sufficient which is not available when object is detected at 100 meters at 65 mph
Solution Approach 1:
The system performs preliminary detection of objects at long ranges using cameras and radar, identifying potential hazards well before they become immediate threats. When a stopped vehicle is detected on the shoulder at long range, the system proactively begins planning and executing lane change maneuvers in advance, ensuring sufficient time and distance are available for safe maneuver completion.
Data Source
AI summary
Aspects of the disclosure relate to controlling a vehicle. For instance, using a camera, a first camera image including a first object may be captured. A first bounding box for the first object and a distance to the first object may be identified. A second camera image including a second object may be captured. A second bounding box for the second image and a distance to the second object may be identified. Whether the first object is the second object may be determined using a plurality of models to compare visual similarity of the two bounding boxes, to compare a three-dimensional location based on the distance to the first object and a three-dimensional location based on the distance to the second object, and to compare results from the first and second models. The vehicle may be controlled in an autonomous driving mode based on a result of the third model.