Autonomous Vehicle Module Scheduling for Sensor Delay Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face errors due to delays in data processing between modules, which can lead to accidents when the logic calculation period exceeds the sensor input period.
Innovation Solution
A control system and method for autonomous vehicles that schedules operations of multiple modules to operate in parallel, using a module scheduling system to monitor and control the threads for data reception, processing, and transmission, thereby reducing processing time and preventing delays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the logic calculation period is extended to perform complicated calculations using multiple sensors, then the processing accuracy and reliability are improved, but the processing time increases causing delays that exceed the sensor input period
Solution Approach 1:
The autonomous driving system is divided into multiple independent modules (sensing module, processing module, control module), each handling specific tasks. This segmentation allows parallel processing of different sensor data and calculation tasks, reducing overall processing time while maintaining calculation accuracy through specialized processing in each module.
Solution Approach 2:
Sensor data is buffered and pre-processed before being fed into the main calculation pipeline. The reception buffer temporarily stores sensing values, and preliminary processing operations prepare data in advance, ensuring that calculation modules receive ready-to-process data without waiting for complete sensor input cycles, thus reducing processing delays.
2Measurement precision
If the sensor input period is reduced to frequently detect nearby objects, then the detection precision and safety are improved, but the processing load increases requiring shorter calculation periods
Solution Approach 1:
Different sensor modules (LiDAR, radar, cameras) and processing tasks are segmented into independent processing units. This allows the system to handle high-frequency sensor inputs from multiple sources simultaneously through parallel processing, maintaining detection precision while managing the increased processing load through distributed computation across multiple modules.
Solution Approach 2:
The system implements continuous processing pipelines where sensor data flows continuously through buffering and processing stages. The reception buffer ensures continuous data availability, and multiple processing modules operate continuously in parallel, maintaining high processing capability that can handle reduced sensor input periods without creating bottlenecks.
3Device complexity
If modules operate sequentially in a traditional pipeline, then the system complexity is reduced, but the processing efficiency decreases causing delays between data reception and transmission
Solution Approach 1:
The system is segmented into distinct functional modules (sensing, processing, control) that can operate independently and in parallel. Each module has dedicated buffers and processing threads, allowing simultaneous execution of multiple operations across modules while maintaining clear module boundaries that manage system complexity through modular design.
Solution Approach 2:
The system transitions from sequential single-threaded processing to multi-threaded parallel processing across multiple modules. By adding the dimension of parallel execution with separate reception threads, main threads, and transmission threads operating simultaneously in different modules, the system achieves higher processing efficiency without proportionally increasing module count.
Data Source
AI summary
A control system of an autonomous vehicle for performing control for effective calculation by parallelizing operations between a plurality of modules of the autonomous vehicle includes the plurality of modules arranged in the autonomous vehicle and configured to perform a connective operation, and a module scheduling system configured to monitor operations of the plurality of modules, to parallelize the plurality of modules using an additional thread, and to control the plurality of modules to aperiodically operate using a pipelining method, and thus, processing capability per hour of a plurality of modules may be optimized by controlling parallel processing of the plurality of modules, and when an entire logic calculation period is longer than a sensor input period, an error due to delay that occurs in a data processing procedure between the modules is overcome.