Automated Driving Perception Model for Multi-Sensor Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle perception systems rely heavily on lidar data, which is expensive and prone to errors due to weather conditions and maintenance issues, and have inefficiencies due to the large number of models required for different tasks, leading to computational overhead and limited scalability.
Innovation Solution
The implementation of an end-to-end perception model (EEPM) that processes and combines features from various sensing modalities like cameras, radars, and lidar, using a shared feature space to improve robustness and efficiency, allowing for flexible training and deployment across different levels of driving automation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If lidar data is used for object detection in autonomous vehicles, then measurement precision is improved, but cost and reliability deteriorate due to expense and susceptibility to weather conditions
Solution Approach 1:
The patent combines multiple sensing modalities (lidar, radar, camera) into a unified end-to-end perception model. This merging allows the system to leverage the high precision of lidar while compensating for its reliability issues through redundant sensing channels that are robust to weather conditions.
Solution Approach 2:
The end-to-end perception model serves multiple functions simultaneously: it processes data from different sensor types, performs object detection, classification, and tracking, and adapts to various weather conditions. This multi-functionality reduces dependency on any single sensor type.
2Adaptability or versatility
If multiple separate models are used for different perception tasks, then adaptability is improved, but device complexity and computational overhead increase
Solution Approach 1:
The patent implements a universal end-to-end perception model that handles multiple perception tasks (detection, classification, tracking) within a single unified architecture. This eliminates the need for separate specialized models while maintaining adaptability through the model's ability to process various sensor inputs and perform diverse output tasks.
Solution Approach 2:
The patent merges multiple task-specific models into a single integrated end-to-end perception system. This consolidation reduces device complexity by eliminating redundant components while preserving task adaptability through the unified model's flexible processing capabilities.
3Adaptability or versatility
If multiple separate models are used for different perception tasks, then task coverage is improved, but productivity deteriorates due to computational overhead
Solution Approach 1:
The patent combines multiple separate perception models into a single end-to-end system that processes sensor data in one unified pipeline. This merging eliminates redundant computational steps and data passing between models, significantly improving processing efficiency while maintaining comprehensive task coverage.
Data Source
AI summary
The described aspects and implementations enable efficient object detection and tracking. In one implementation, disclosed is a method and a system to perform the method, the system including the sensing system configured to obtain sensing data characterizing an environment of the vehicle. The system further includes a data processing system operatively coupled to the sensing system and configured to process the sensing data using a first (second) set of neural network (NN) layers to obtain a first (second) set of features for a first (second) region of the environment, the first (second) set of features is associated with a first (second) spatial resolution. The data processing system is further to process the two sets of features using a second set of NN layers to detect a location of obj ect(s) in the environment of the vehicle and a state of motion of the object(s).