Evaluation Of RISC-? New Embedded Platform In Vehicle Behavior Decision Algorithm

The behavioral decision algorithm in the autonomous driving system is a kind of computationally intensive artificial intelligence application,which puts forward new requirements on the design of the underlying hardware.As an emerging open source instruction set architecture,RISC-? has broad prospects in the design of CPUs and accelerators,and is expected to be used in the hardware platform of autonomous driving systems.However,the majority of assisted driving systems have not yet been practiced and evaluated on the RISC-? platform.In response to the above problems,combined with the characteristics of the vehicle embedded platform,the evaluation of the RISC-? embedded platform in the behavioral decision algorithm is designed.The transplantation and evaluation consists of the following parts: 1)In the evaluation design part,several evaluation indicators are designed according to the characteristics of the RISC-? platform in the embedded environment,and the influence of these indicators on the algorithm is discussed;2)The extended instruction set evaluation is mainly for evaluating and testing the meaningful RISC-? extended instruction set in embedded scenarios,analyzing the impact of the extended instruction set on program operation,and the evaluation results are used to guide the choice of instruction set in hardware design;3)Storage resource evaluation is mainly aimed at evaluating and testing the hardware resource usage including registers and memory in the embedded environment,and is used to guide hardware design and improve resource utilization.The above experiments are carried out on the RISC-? instruction simulator,and the compression instruction set and storage resources are tested separately.The experimental results show that the program with the compressed instruction set as the target architecture reduces the static instruction volume by about 21.24% and the dynamic instruction length by 17.1%.The registers used during program operation are concentrated in 14 registers,and the frequency of other registers is extremely low;at the same time,the memory required for program operation is small,and the size of the data set has little impact on the required memory.The data shows that the vehicle behavior decision algorithm has good adaptability on the RISC-? embedded platform that supports the compression and expansion instruction set,and the demand for storage resources is low.