Snake robots are ideally suited to highly confined environments because their small cross-sections and highly redundant kinematics allow them to enter and move through tight spaces with a high degree of dexterity. Despite these theoretical advantages, snake robots also pose a number of practical challenges that have limited their usefulness in the field. These challenges include the need to coordinate a large number of degrees of freedom, decreased system reliability due to the serial nature of the robot’s design, and the complex interaction of the robot’s shape with the world.
This thesis makes progress towards addressing these issues with two main areas of contribution. In the first part, we provide tools for assisted autonomy in snake robots. To provide intuitive high-level autonomous behaviors, we extend our lab’s existing gait-based control framework to develop gait-based compliant control. To reliably and accurately sense the robot’s pose and shape we present new techniques for robust state estimation that leverage the redundancies in the distributed sensing capabilities of our group’s articulated snake robots.
To demonstrate these contributions in a practical application, we use them to enable a snake robot to navigate a real-world underground pipe network. One of the most limiting characteristics of our snake robots (and robots in general) is the inability to precisely sense and control the torques and forces of their actuators. As such, the second part of this thesis focuses on the design and control of a new series-elastic snake robot that incorporates a high performance series-elastic actuator (SEA) and torque control. After describing the novel design of the SEA, we discuss our perspective on how to incorporate torque control and series elasticity into snake robots. Finally, we demonstrate prototypes of new low impedance motions for snake robots. These motions provide significantly more affordance to obstacles and unstructured terrain, and open a new avenue of research for snake robot locomotion.
Howie Choset (Chair)
Art Kuo (University of Michigan)
lyonsmuth [atsymbol] cmu.edu