Dynamic Bipedal Walking while Carrying an Unknown Time-Varying Load

In recent years, there has been tremendous interest in employing legged and humanoid robots for dangerous missions in disaster and rescue scenarios, requiring robots to operate swiftly and stably while dealing with high levels of uncertainty and large external disturbances. One such practical example is using a legged robot to carry and manipulate large and heavy objects whose mass is not known a priori. Most prior work only considers constant static loads. For instance, traditional controllers for dynamic bipedal walking, such as in [2], fails when the unknown mass exceeds 15% of the robot mass. Results in [1] demonstrate robust adaptive control to adapt to an unknown load of 16 % of robot weight. The problem of carrying large loads while maintaining dynamic walking on a nonlinear, underactuated and hybrid dynamical system is challenging. The limitation of current research, as well as the demand of practical requirement, motivates our research on using adaptive control for dynamic bipedal walking while carrying large unknown time-varying load.