Overviewed technical vision unit usage for solving walking robot's navigational task and, also, determination of the robot's angular orientation relative to observed scene. Shown, that single moving camera can be used for a stereo image acquisition with future computation of optical flow, instead of rigid stereo-camera module. That allows to minimize computational requirements of the data transfer unit. Shown basic overview of the robot's kinematic model and walking principle. Described general task of optical flow computation. Provided overview of optical flow computation principle and results of quantitative comparison of nine different techniques, including instances of differential methods, region-based matching energy-based and phase-based techniques namely those of Horn and Schunck, Lucas and Kanade, Uras et al., Nagel, Anandan, Singh, Heeger, Waxman et al. and Fleet and Jepson. Shown Lucas-Kanade approach for optical flow computation technique. Provided general principle and implementation for obstacle avoidance basing on optical flow technique, provided quantitative results implemented in current robot. Shown next stages of obstacle avoidance: obstacle detection by means of optical flow relative velocities threshold ; distance to the obstacle determination by means of known stereo-image parameters; obstacle relative speed determination by means of optical flow of obstacle and it's direction and position; obstacle size, also by means of optical flow and known configuration of scene, and relative to robot angular position by means of optical eigenvectors computation with future differentiation and solving Poisson's equation.