The invention belongs to the technical field of path planning, and discloses an unmanned driving dynamic path
planning method and
system based on multi-
source data fusion, and the method comprises the steps: collecting an ice and
snow pavement
friction coefficient, a curve curvature and an obstacle
point cloud, constructing a sensor confidence
coefficient matrix, generating a fused
semantic map, and constructing a dynamic environment semantic model. Outputting a real-time
friction coefficient field and a risk thermodynamic map layer; the roadside unit broadcasts coordinates of opposite vehicles in a blind area of a curve to a vehicle end, constructs an ice and
snow pavement offset
crowdsourcing map, and generates a global-local fusion topology; fusing the real-time
friction coefficient field and the global-local fusion topology to generate a smooth trajectory set, and further generating a risk optimal path
instruction set; a
steering angle and torque instruction is decomposed, positioning drift is compensated in real time, and a
normal mode for updating the
vehicle positioning state and a degradation mode when the
millimeter wave radar fails are constructed; and generating execution logs and health state vectors, and aggregating the execution logs and the health state vectors of multiple vehicles to form closed-loop iterative update.