Visual Saliency Perception for Scenario-Specific Autonomous Driving Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in effectively managing operational scenarios, particularly in navigating complex transportation networks and detecting external objects, due to limitations in perception and control systems, which can lead to safety and efficiency issues.
Innovation Solution
An autonomous vehicle operational management system that incorporates a perception unit using deep learning for visual saliency perception, processing sensor data to detect external objects and generate saliency information, and instantiates scenario-specific control modules to control vehicle actions, enabling safe and efficient navigation through vehicle transportation networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sensor data processing is used, then the system is simpler, but object detection precision and safety are insufficient
Solution Approach 1:
The patent introduces an operational scenario module as an intermediary between sensor data acquisition and control actions. This module processes sensor data through multiple layers (data acquisition, processing, analysis, scenario identification) to generate scenario-specific control actions, thereby improving detection precision while managing system complexity through structured intermediate processing stages
Solution Approach 2:
The perception system is segmented into distinct functional modules: sensor data acquisition module, operational environment monitor, scenario-specific operational control evaluation module instances, and autonomous vehicle operational management controller. Each module handles specific tasks independently, improving overall detection precision through specialized processing while organizing complexity into manageable segments
2Adaptability or versatility
If generic control modules are used, then the system is simpler, but adaptability to different operational scenarios is reduced
Solution Approach 1:
The system dynamically instantiates scenario-specific operational control evaluation module instances based on detected operational scenarios. The autonomous vehicle operational management controller adapts the control architecture in real-time by creating, modifying, or removing module instances according to the current operational context, thereby achieving high adaptability while managing complexity through dynamic reconfiguration
Solution Approach 2:
The system changes operational parameters by instantiating different control module instances with scenario-specific parameters. Each operational scenario triggers the creation of control modules configured with parameters optimized for that specific scenario (e.g., pedestrian crossing, intersection, parking), allowing the system to adapt to diverse conditions without requiring a completely different control architecture for each case
3Reliability
If real-time perception and control are implemented, then safety and responsiveness improve, but computational load and system complexity increase
Solution Approach 1:
The system performs preliminary actions by pre-defining operational scenario templates and control module structures. When an operational scenario is detected, the system instantiates pre-configured control module instances rather than creating control logic from scratch, thereby achieving real-time responsive control while reducing computational burden through preparatory structuring of control architectures
Solution Approach 2:
The autonomous vehicle operational management system implements continuous feedback loops where the operational environment monitor constantly monitors sensor data, identifies operational scenarios, and triggers appropriate control module instances. The system receives feedback from control actions and continuously adjusts operational parameters, ensuring high reliability and safety through real-time closed-loop control while managing complexity through structured feedback processing
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Autonomous vehicle operational management with visual saliency perception control may include operating a perception unit and an autonomous vehicle operational management controller. Operating the perception unit may include generating external object information based on image data received from image capture units of the vehicle and saliency information received from the autonomous vehicle operational management controller. Operating the autonomous vehicle operational management controller may include identifying a distinct vehicle operational scenario based on the external object information, instantiating a scenario-specific operational control evaluation module instance, receiving a candidate vehicle control action from a policy for the scenario-specific operational control evaluation module instance, and controlling the autonomous vehicle to traverse a portion of the vehicle transportation network in accordance with the candidate vehicle control action, wherein the portion of the vehicle transportation network includes the distinct vehicle operational scenario.