Snakes on a Plane: Modeling, Estimation and Locomotion for Planar Snake robots in Cluttered Environments

Abstract

Snake robots are mechanisms designed to mimic biological snakes, and aspire to inherit the robustness and stability of their biological counterparts. As of yet, this is a largely unrealized potential. This thesis aims to explore various topics in modeling, state estimation and control pertaining to snake robot locomotion in cluttered environments. The thesis is divided into three constituent parts, each exploring one topic. Chapter 2 concerns the design and development of the Boa snake robot, a next-generation sensor-driven snake robot research platform. In addition to embedded gyroscopes and accelerometers in each link, the Boa includes a constraint force measurement system capable of measuring the interaction forces acting between adjacent links in the body of the robot. In the Boa, this system has been improved and made less prone to mechanical failure. This chapter includes the design and development of mechanical and mechatronic parts for the robot, on-board electronics, sensor systems, and software. Chapter 3 is dedicated to modeling and estimation leveraging the on-board sensors in the Boa snake robot. Here, we aim to show that intrinsic sensor data can be used to create meaningful estimates of the robots’ interaction with its environment in the form of contact point and contact force estimates. We show how this data can also be used to create a simple model of the robots’ otherwise complex dynamics. Finally, we show how the Unscented Kalman Filter (UKF) can be applied to perform real-time estimates of the manipulator Jacobian matrix during locomotion. This method is shown to outperform a similar method adapted from soft robotics in both statistical properties and execution time. Chapter 4 is dedicated to the theory of form closure, a mathematical concept that has found widespread use in computing grasps for prehensile robots to grip objects with different geometry. In this part, we present a tutorial on the foundational theory of form closure, and expand the theory to the field of snake robots. We show how form closure can be used to identify the form closed region of the robots’ configuration space. In this region, the robot can be modeled a fully actuated system, which is shown to be beneficial to modeling, control and locomotion.