By invoking a cut of satellite image as an as-is representation of local terrain, roadway pattern can be segmented and extended towards possible destinations prior to physical arrival. Guided by the roadway pattern to be projected into associated scene image, in this paper, stochastic perception-control coupling is introduced for anticipative maneuvering through naturally complex scene. In this scheme, the perception process generates boundary probability and matches with fractal covering of object free space. Such a saccadic guidance signal is smoothed via Gaussian interpolation mechanism and transfered to the control process where trajectory of vehicle mechanism is regulated as a stochastic process subjected to human's unexpectable decision making. The validity of stochastic trajectory generation is verified via simulation studies.
展开▼
