新一代信息技術
Driving intention inferring method optimized by full transfer learning based long short-term memory
The present invention belongs to the technical field of active safety of electric vehicles, and in particular to a driving intention inferring method optimized by a full transfer learning based long short-term memory (LSTM) network. The driving intention inferring method includes: step 1, building a multi-variable fractional gray model of a driving intention; step 2, determining whether a source domain and a target domain have similarities; step 3, designing and fully transferring an LSTM network; and step 4, optimally computing fractional orders, to determine the driving intention. The present invention infers the driving intention directly from a road surface condition, and has little interference information and high precision; and with introduction of an absolute gray correlation degree, probability computation of a large amount of data is omitted, thereby greatly reducing computation burden.
長春工業大學
授權發明
