The invention belongs to the field of embedded
intelligent computing, and provides a heterogeneous fusion embedded
intelligent computing implementation method. The method comprises the steps of determining
algorithm characteristics of a required core intelligent
algorithm by analyzing task characteristics of intelligent tasks in an equipment
embedded system OODA full task chain, determining and selecting a core intelligent
algorithm of a corresponding category according to the required algorithm characteristics, wherein the core intelligent algorithm is divided into a knowledge-driving algorithm, an intelligent optimization algorithm and a
deep learning algorithm, scheduling through
application service management, mapping the selected core intelligent algorithm to a corresponding softwareoperation framework in a
system platform
service layer, wherein all the
software running frameworks are loaded into an
adaptive hardware module when the
system is started, and selecting a hardware module through resource service management to load the core intelligent algorithm. According to the method, efficient, flexible and universal
intelligent computing support service can be provided for equipment OODA autonomous tasks, and execution of the equipment autonomous tasks is accelerated.