Terrain Classification Strategy of a Quadruped Robot for Gait Transition and Adaptation in a Field Terrain > Journal

본문 바로가기
Advanced Science Letters 국제 저널 VOL.19, NO.1, PP.21-26

Terrain Classification Strategy of a Quadruped Robot for Gait Transition and Adaptation in a Field Terrain

Seungnam Yu, Kwangjin Ko, Kisung Kim, Wansoo Kim
논문 보기 관리자 · 조회 897 · 2013-12-31
ABSTRACT

A terrain sensing and classification technique for a legged-type field robot is one of the most important issues that can influence the robot's performance because the field conditions that can be encountered by the robot are extremely diverse. On a flat surface, for instance, the robot might walk normally but it would have trouble walking on irregular terrain. This research starts from the idea that the robot walks while recognizing its current terrain surface. If the robot can treat and select the proper gait against the terrain for every instant, the burden and complexity of the classical control algorithm for the robot while walking can be reduced. In this paper, a quadruped robot is developed to apply and verify the designed gait. Additionally, to develop the terrain recognition algorithm, a multi-purposed 1-legged robot system is proposed. This system can reproduce the designed gait of a 4-legged system on the terrain. In the test-bed of the 1-legged system, four kinds of considered field conditions are applied for the terrain classification experiment. Especially, a load cell mounted on the leg and torque sensors located on the joints, respectively to measure the ground reaction force and joint torque while walking. The measured data is then characterized using the statistical method and those average values are learned using the back propagation neural network (BPNN) algorithm. Finally, we verified the proposed method for gait classification and applied to the 4-legged robot to test the real-time gait transition performance with the design of two types of gaits.