Hard-Isolated Sensor Data Access System for Autonomous Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face performance loss due to resource-intensive virtual machine isolation and lack of real-time execution capabilities, posing safety risks from asynchronous tasks and interrupts.
Innovation Solution
A sensor data access system with hard-isolated hardware domains and inter-processor communication, eliminating the need for virtual machines to share sensor data, utilizing real-time processors and interrupt modules for secure and efficient data access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If virtual machine technology is used to isolate autonomous driving tasks, then system isolation is achieved, but resource consumption increases and performance loss occurs
Solution Approach 1:
The system is divided into multiple hardware domains (sensor hardware domain, first hardware domain, second hardware domain) that are hard-isolated from each other. Each domain operates independently with its own processor and memory space, achieving system isolation without virtual machine overhead. The sensor hardware domain contains sensor modules (ultrasonic radar, millimeter wave radar, camera) that are physically separated from computation domains.
Solution Approach 2:
An inter-processor communication module serves as an intermediary between the sensor hardware domain and computation hardware domains. This module enables controlled data exchange and sensor sharing between domains through defined interfaces, allowing efficient communication while maintaining hard isolation boundaries. The intermediary handles data requests from multiple hardware domains without requiring virtual machine abstraction.
2Reliability
If multiple virtual machines are used for different autonomous driving tasks, then task isolation is achieved, but performance loss is exacerbated
Solution Approach 1:
The system divides autonomous driving tasks across separate hardware domains rather than using multiple virtual machines. Each hardware domain (first hardware domain, second hardware domain) has dedicated processors that execute tasks without virtualization overhead. This segmentation eliminates the performance penalty of context switching and resource virtualization while maintaining task isolation through physical separation.
Solution Approach 2:
The patent replaces the software-based virtual machine mechanism with a hardware-based domain isolation architecture. Instead of using virtual machine monitors to manage task isolation, the system uses physically separated hardware domains with dedicated processors and memory spaces, eliminating the computational overhead of virtual machine management and improving overall system performance.
3Ease of operation
If traditional operating systems are used in autonomous driving systems, then ease of operation is maintained, but real-time execution capability is insufficient
Solution Approach 1:
The system separates real-time critical tasks from non-real-time tasks by placing them in different hardware domains. The first hardware domain with its dedicated processor handles real-time autonomous driving tasks that require deterministic execution, while other domains handle less time-critical functions. This segmentation allows real-time performance without requiring a real-time operating system for the entire system.
Solution Approach 2:
The inter-processor communication module acts as a mediator that handles communication between domains with different timing requirements. It manages data exchange and coordinate between real-time and non-real-time domains, allowing traditional operating systems to be used in non-critical domains while maintaining real-time performance in critical domains through hardware isolation.
4Adaptability or versatility
If sensor data is shared through virtual machines, then sensor sharing is achieved, but system complexity increases
Solution Approach 1:
The system places sensor modules (ultrasonic radar sensor, millimeter wave radar sensor, camera sensor) in a dedicated sensor hardware domain that is physically separated from computation domains. This segmentation allows multiple hardware domains to access sensor data through the inter-processor communication module without requiring complex virtual machine configurations. Each domain can independently request and process sensor data relevant to its function.
Solution Approach 2:
The inter-processor communication module serves as an intermediary that manages sensor data sharing between the sensor hardware domain and multiple computation hardware domains. It handles data requests from different domains, performs necessary data formatting and routing, and maintains data consistency without requiring virtual machine abstraction layers. This intermediary approach simplifies the overall system structure compared to virtual machine-based sensor sharing.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
Present application relates to a sensor data access system, a sensor data access system comprising a plurality of domains including plurality of hardware domains and at least one sensor hardware domain on a system-on-chip. Each domain is hard-isolated from another domain. The sensor hardware domain is connected with the plurality of hardware domains respectively through inter-processor communication module, the sensor hardware domain has at least one sensor module. The sensor module is used for receiving sensor data for autonomous driving. The plurality of the hardware domains include a first hardware domain and a second hardware domain provided on the same system-on-chip. The first hardware domain acquires sensor data from the sensor hardware domain, and the second hardware domain also acquires sensor data from the sensor hardware domain. Sensor sharing can be achieved without importing virtual machine technology, thereby reducing the cost of implementing autonomous driving.