Gait Optimization of a Quadruped Robot Using Evolutionary Computation

Evolutionary Computation (EC) has strengths in terms of computation for gait optimization. However, conventional evolutionary algorithms use typical gait parameters such as step length and swing height, which limit the trajectory deformation for optimization of the foot trajectory. Furthermore, the quantitative index of fitness convergence is insufficient. In this paper, we perform gait optimization of a quadruped robot using foot placement perturbation based on EC. The proposed algorithm has an atypical solution search range, which is generated by independent manipulation of each placement that forms the foot trajectory. A convergence index is also introduced to prevent premature cessation of learning. The conventional algorithm and the proposed algorithm are applied to a quadruped robot; walking performances are then compared by gait simulation. Although the two algorithms exhibit similar computation rates, the proposed algorithm shows better fitness and a wider search range. The evolutionary tendency of the walking trajectory is analyzed using the optimized results, and the findings provide insight into reliable leg trajectory design.

[1]  Masahiro Fujita,et al.  Autonomous evolution of dynamic gaits with two quadruped robots , 2005, IEEE Transactions on Robotics.

[2]  Wei Wang,et al.  Stable quadruped walking with the adjustment of the center of gravity , 2013, 2013 IEEE International Conference on Mechatronics and Automation.

[3]  Masakatsu G. Fujie,et al.  Foot trajectory for a quadruped walking machine , 1990, EEE International Workshop on Intelligent Robots and Systems, Towards a New Frontier of Applications.

[4]  Ludovic Righetti,et al.  Pattern generators with sensory feedback for the control of quadruped locomotion , 2008, 2008 IEEE International Conference on Robotics and Automation.

[5]  Zdenko Kovacic,et al.  Multiobjective locomotion optimization of quadruped robot with different 2DOF configurations of actuated spine , 2016, 2016 24th Mediterranean Conference on Control and Automation (MED).

[6]  Tomohiro Yamaguchi,et al.  Free-gait of quadruped robots using neural networks for determining the order of swing leg , 2004, 2004 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (IEEE Cat. No.04CH37566).

[7]  L. M. Day,et al.  Interspecific scaling of the morphology and posture of the limbs during the locomotion of cats (Felidae) , 2007, Journal of Experimental Biology.

[8]  Andrew Y. Ng,et al.  A control architecture for quadruped locomotion over rough terrain , 2008, 2008 IEEE International Conference on Robotics and Automation.

[9]  Tao Mei,et al.  Design of Quadruped Robot Based Neural Network , 2007, ISNN.

[10]  Chao Ma,et al.  Learning a Faster Locomotion Gait for a Quadruped Robot with Model-Free Deep Reinforcement Learning , 2019, 2019 IEEE International Conference on Robotics and Biomimetics (ROBIO).

[11]  Jun Wu,et al.  A Multi-module Controller for Walking Quadruped Robots , 2019, Journal of Bionic Engineering.

[12]  Thomas Röfer,et al.  Evolutionary Gait-Optimization Using a Fitness Function Based on Proprioception , 2004, RoboCup.

[13]  Huosheng Hu,et al.  Parameter optimisation of an evolutionary algorithm for on-line gait generation of quadruped robots , 2003, IEEE International Conference on Industrial Technology, 2003.

[14]  Roland Siegwart,et al.  Toward Combining Speed, Efficiency, Versatility, and Robustness in an Autonomous Quadruped , 2014, IEEE Transactions on Robotics.

[15]  Jong Hyeon Park,et al.  Trajectory optimization with GA and control for quadruped robots , 2009 .

[16]  Jong Hyeon Park,et al.  Ellipse-based leg-trajectory generation for galloping quadruped robots , 2008 .

[17]  Manuela M. Veloso,et al.  An evolutionary approach to gait learning for four-legged robots , 2004, 2004 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (IEEE Cat. No.04CH37566).

[18]  Xiaolei Han,et al.  Trot Gait Design and CPG Method for a Quadruped Robot , 2014 .

[19]  Peter Stone,et al.  Machine Learning for Fast Quadrupedal Locomotion , 2004, AAAI.

[20]  Atil Iscen,et al.  Sim-to-Real: Learning Agile Locomotion For Quadruped Robots , 2018, Robotics: Science and Systems.

[21]  M. Nili Ahmadabadi,et al.  Modeling, control and gait design of a quadruped robot with active spine towards energy efficiency , 2015, 2015 3rd RSI International Conference on Robotics and Mechatronics (ICROM).

[22]  Auke Jan Ijspeert,et al.  Kinematic primitives for walking and trotting gaits of a quadruped robot with compliant legs , 2014, Front. Comput. Neurosci..

[23]  Peter Stone,et al.  Policy gradient reinforcement learning for fast quadrupedal locomotion , 2004, IEEE International Conference on Robotics and Automation, 2004. Proceedings. ICRA '04. 2004.

[24]  Mete Kalyoncu,et al.  Optimal Tuning of PID Controller Using Grey Wolf Optimizer Algorithm for Quadruped Robot , 2018 .

[25]  M. O. Tokhi,et al.  Training neural networks: backpropagation vs. genetic algorithms , 2001, IJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No.01CH37222).

[26]  R. Alexander,et al.  The relationship between limb‐segment proportions and joint kinematics for the hind limbs of quadrupedal mammals , 2002 .

[27]  Jie Yan,et al.  A Review of Gait Optimization Based on Evolutionary Computation , 2010, Appl. Comput. Intell. Soft Comput..

[28]  William T. B. Uther,et al.  Automatic Gait Optimisation for Quadruped Robots , 2003 .

[29]  Darwin G. Caldwell,et al.  Dynamic trot-walking with the hydraulic quadruped robot — HyQ: Analytical trajectory generation and active compliance control , 2013, 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems.

[30]  Vítor Matos,et al.  Gait transition and modulation in a quadruped robot: A brainstem-like modulation approach , 2011, Robotics Auton. Syst..

[31]  L. Ren,et al.  Review and Classification of Bio-inspired Algorithms and Their Applications , 2020 .

[32]  Kenneth O. Stanley,et al.  Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternative for Training Deep Neural Networks for Reinforcement Learning , 2017, ArXiv.