Heterogeneous Compute Architecture for Real-Time Autonomous Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automotive systems face challenges in managing the proliferation of Embedded Electronic Control Units (ECUs) due to their extensive array of functions, requiring improved manageability, real-time performance, and safety in autonomous driving applications, which existing solutions fail to address effectively.
Innovation Solution
A heterogeneous compute architecture (HCA) for hardware/software co-design is introduced, featuring scalable processors, flexible networking, and benchmarking tools, allowing for rapid architecting and evolving of vehicle system architectures, integrating FPGA, GPU, or ASIC solutions via PCIe or other interconnects, and providing a platform for autonomous driving system development that meets stringent safety, cost, and power efficiency requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional ECUs are used to cover extensive array of functions, then functional versatility is improved, but device complexity and manageability deteriorate
Solution Approach 1:
The patent segments the monolithic ECU system into a distributed network of heterogeneous compute units (HCUs) with specialized functions. Each HCU handles specific autonomous driving tasks (sensing, perception, planning, control) independently, reducing overall system complexity while maintaining functional versatility through modular architecture.
Solution Approach 2:
The patent implements universal communication interfaces and standardized protocols (CAN bus, Ethernet, PCIe) that enable different types of compute units to interoperate seamlessly. This universal interface layer allows the system to maintain versatility across diverse functional components while simplifying integration and management.
2Adaptability or versatility
If more ECUs are added to cover more functions, then functional capability is improved, but system manageability and real-time performance deteriorate
Solution Approach 1:
The patent divides autonomous driving functions into segmented processing stages across specialized compute units: sensor data acquisition units, perception processing units, planning units, and control units. This segmentation enables real-time performance by allowing parallel processing of different function groups while maintaining reliable inter-unit communication through standardized interfaces.
Solution Approach 2:
The patent implements dynamic task allocation and load balancing mechanisms where compute units can adaptively assign processing tasks based on real-time system conditions and workload demands. This dynamic resource management ensures real-time performance requirements are met while maintaining functional versatility through flexible task distribution.
3Productivity
If heterogeneous compute architecture is implemented, then processing power and real-time performance are improved, but hardware complexity increases
Solution Approach 1:
The patent employs universal communication protocols and standardized interface layers (CAN bus, Ethernet, PCIe, SPI, I2C) that enable heterogeneous compute units with different processing capabilities to interoperate seamlessly. This universal interface strategy allows the system to leverage diverse hardware for enhanced processing power while maintaining manageable hardware complexity through standardized connectivity.
Solution Approach 2:
The patent introduces communication control units and gateway components that act as intermediaries between heterogeneous compute units with different architectures. These intermediary components translate and manage data flow between diverse hardware platforms, enabling high processing power through hardware diversity while reducing overall system complexity through centralized communication management.
Data Source
AI summary
Methods and apparatus relating to heterogeneous compute architecture hardware/software co-design for autonomous driving are described. In one embodiment, a heterogeneous compute architecture for autonomous driving systems (also interchangeably referred to herein as Heterogeneous Compute Architecture or “HCA” for short) integrates scalable heterogeneous processors, flexible networking, benchmarking tools, etc. to enable (e.g., system-level) designers to perform hardware and software co-design. With HCA system engineers can rapidly architect, benchmark, and/or evolve vehicle system architectures for autonomous driving. Other embodiments are also disclosed and claimed.