Autonomous Vehicle Compute Redundancy for Assured Output Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current functional safety standards, such as ISO 26262, are incompatible with the implementation of autonomous vehicle functionality and are inflexible, making it challenging to assure the processing operations of an autonomy computing system.
Innovation Solution
An autonomous vehicle compute architecture that includes a plurality of functional circuits and monitoring circuits, allowing for the evaluation of output consistency across multiple processing paths to generate assured outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If functional safety standards (ISO 26262) are strictly implemented, then hardware and software standards are established to assure proper functionality, but the standards become incompatible with autonomous vehicle functionality and overly inflexible
Solution Approach 1:
The system divides the autonomous vehicle computing system into multiple independent functional circuits (first functional circuit, second functional circuit) that can be individually validated. Each circuit processes sensor data independently to generate outputs, allowing modular safety verification without requiring complete system redesign to meet functional safety standards.
Solution Approach 2:
The system creates redundant copies of functional circuits that perform identical processing tasks. The first functional circuit and second functional circuit both process sensor data to generate motion plans, allowing cross-validation of outputs. This copying approach enables safety assurance through comparison while maintaining autonomous vehicle functionality.
2Reliability
If multiple functional circuits are used to generate outputs for validation, then output consistency can be evaluated, but system complexity increases
Solution Approach 1:
The system merges the validation function into the existing compute architecture by having functional circuits share common components such as sensor data inputs, processing algorithms, and output interfaces. This consolidation allows multiple functional circuits to be implemented without proportionally increasing overall system complexity, as they leverage shared resources.
Solution Approach 2:
The functional circuits are designed with universal characteristics, using the same sensor data inputs, processing pipelines, and output formats. This multi-functionality allows the same hardware and software components to serve multiple circuits simultaneously, reducing the complexity overhead of implementing redundant validation paths.
3Reliability
If functional circuits process sensor data to generate motion plans, then autonomous vehicle control is achieved, but processing time and computational load increase
Solution Approach 1:
The system performs preliminary processing by having functional circuits process sensor data in parallel to generate multiple candidate motion plans simultaneously. Rather than sequentially validating each plan, the system pre-computes multiple outputs and then rapidly compares them, reducing the critical validation time while maintaining thoroughness.
Solution Approach 2:
The system replaces complex sequential validation mechanisms with direct output comparison between functional circuits. Instead of using elaborate verification procedures to validate motion plans, the system simply compares the outputs of parallel circuits and selects the consistent result, significantly reducing processing time while maintaining reliability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems and methods are directed to a method for assured autonomous vehicle compute processing. The method can include providing sensor data to first and second functional circuitry of an autonomy computing system. The first and second functional circuitry can be configured to generate first and second outputs associated with a first autonomous compute function. The method can include generating, by the first and second functional circuitry in response to the sensor data, first and second output data associated with the first autonomous compute function. The method can include generating, by monitoring circuitry of the autonomy computing system, comparative data associated with differences between the first output data and the second output data. The method can include generating one or more vehicle control signals for the autonomous vehicle based at least in part on the comparative data.