Real-time obstacle avoidance using central flow divergence and peripheral flow

Abstract

The lure of using motion vision as a fundamental element in the perception of space drives this effort to use flow features as the sole cues for robot mobility. Real-time estimates of image flow and flow divergence provide the robot's sense of space. The robot steers down a conceptual corridor, comparing left and right peripheral flows. Large central flow divergence warns the robot of impending collisions at "dead ends. " When this occurs, the robot turns around and resumes wandering. Behavior is generated by directly using flow-based information in the 2-D image sequence; no 3-D reconstruction is attempted. Active mechanical gaze stabilization simplifies the visual interpretation problems by reducing camera rotation. By combining corridorfollowing and dead-end deflection, the robot has wandered around the lab at 30 cm/s for as long as 20 minutes without collision. The ability to support this behavior in real-time with current equipment promises expanded capabilities as computational power increases in the future.

End-to-End Depth From Motion With Stabilized Monocular Videos

“…One can note that rotation compensation was already used with fish-eye camera in order to have a more direct link between flow and depth. Another work (Coombs et al, 1998) also demonstrated that basic obstacle avoidance could be achieved in cluttered environments such as a closed room.…”

Ground-plane classification for robot navigation: Combining multiple cues toward a visual-based learning system

“…This method is invariant under many assumptions inherently associated with appearance-based methods. Although the qualitative use of optical-flow has had good success for navigation through techniques such as balancing optical-flow fields [15], we focus on those techniques which are beneficial toward the task of image classification. The idea of ground-plane detection using optical-flow is not new [16] [17].…”

Autonomous exploration using rapid perception of low-resolution image information

“…Traditional methods of computing TTC [1,9] require computing the divergence of the estimated optical flow, which is not only computationally intensive but, more importantly, requires a significant amount of texture in the scene. To overcome these problems, Horn et al [14] have recently described a direct method to determine the time-to-collision using image brightness derivatives (temporal and spatial) without any calibration, tracking, or optical flow estimation.…”