| This thesis addresses the dynamics, gait generation and motion planning of a class of locomotion systems called modular locomotion systems that consist of a central base module with locomotion modules like legs, powered wheels or passive wheeled skate modules attached to ports on the base. These systems, by virtue of their often unconventional modes of locomotion, offer advantages in performing tasks where traditional wheeled and legged systems often fail, for example crawling through pipes, skating on ice, or climbing fences.; This thesis makes multiple contributions to the study of modular locomotion systems in enumeration, kinematic and dynamic analysis and motion planning. We present a scheme to enumerate the different robot configurations possible given a set of locomotive modules. For kinematic systems, we present a motion planning algorithm that incorporates the effect of intermittent contact of leg modules. A dynamic analysis is carried out for systems with multiple nonholonomic constraints using both the Lagrange D'Alembert equations of motion and a process of Lagrangian reduction to define the connection that relates the group motion of the robot to the individual motion of the modules (shape inputs). We show how nonholonomic impacts, which occur when passive wheel modules make or break contact with the ground, can be incorporated into this analysis.; We study the dynamics, gait generation and motion planning for three dynamic systems with an unconventional actuation scheme and multiple nonholonomic constraints. We present simulation and experimental results to show how simple for ward and rotary gaits for the ROLLERBLADER, a robot with two degree of freedom legs with passive wheels mounted on the ends, can be composed to achieve motion planning. We also present a rollerblading gait that attempts to mimic human rollerblading motion. We show how the R OBOTRIKKE, a novel single input experimental system, can be controlled using a periodic input and visual feedback from an overhead camera to track a straight line trajectory. We present simulation results for controllers that balance a bicycle without pedals, propel it forwards and allow simple trajectory tracking once the bicycle reaches higher speeds. |
