The application discloses a kind of man-
machine interaction risk prediction and adaptive safety avoidance methods, systems and media for automatic
plant vehicle, first through multi-source sensing synchronous acquisition
plant vehicle working condition and
plant environment data, rely on pre-constructed plant
semantic map matching
pedestrian space attribute;Solving motion parameters to
pedestrian multi-frame tracking, generate multi-probability prediction trajectory in combination with plant topology;Again, segmentation sensor
occlusion area, infer potential
pedestrian risk in blind area based on multi-feature Sigmoid model;Fusion explicit pedestrian, blind area pedestrian, plant semantic component constructs space-time man-
machine interaction risk field;In combination with vehicle speed, load, trajectory uncertainty Real-time generation dynamic
safety zone;According to risk overlap index Implementation four-level hierarchical control, and supporting sensor
abnormality three-level safety degradation strategy.The application breaks through the defect of traditional fixed safety threshold, only passive
obstacle avoidance, can predict pedestrian multiple motion behaviors in advance, early warning blind area sudden pedestrian risk, applicable to AGV, unmanned fork
truck and other man-
machine mixed plant scene.