This dissertation focuses on modeling and sensing for miniature devices characterized by mixed elastic and rigid body motion with contact. Understanding dynamics of miniature devices, including microelectromechanical systems (MEMS) and related small-scale prototypes, is valuable for improving device design and helping develop estimation and control algorithms for better performance. This work particularly emphasizes dynamic modeling of small-scale devices that depend on contact interactions between structures to generate continuous or repeated motion, and thus modeling will incorporate contact dynamics within devices featuring both compliant and rigid components. The first task is to model the dynamics of a magnetoelastic micro-motor. The magnetoelastic micro-motor has characteristics of large payload capacity and feasibility of bidirectional motion. These characteristics suggest it as a good candidate for gyroscope calibration in the field as a type of miniature rate table. This task proposes a dynamic model including both a compliant model for the stator and bidirectional rigid body motion model for the rotor, with a capacitive sensing design to precisely track the motor motion. The contact interactions between the stator and rotor are the key feature of the micro-motor actuation, so the influence of this interaction is modeled for a better understanding of the micro-motor behavior. Stochastic bouncing motion of the magnetoelastic rotor is not ideal for a controlled motion, such that the dynamic model helps compensate for this limitation of magnetoelastic actuation. Experimental comparison validates the dynamic model for a reproduction of major micro-motor dynamic features. Capacitive sensing based on this work is predicted to compensate for off-axis motion. The second task is to model the nonlinear hybrid dynamics of a piezoelectric walking micro-robot. The piezoelectric micro-robot is suitable for integration with control and power systems because a relatively large payload can be supported at low power. However, piezoelectric actuation usually provides a relatively short actuation stroke, and to overcome this drawback, piezoelectric actuators are often operated near their resonance. Therefore, the dynamics of the piezoelectric micro-robot is formulated using a resonance-based model for its compliant leg mechanisms and rigid chassis motion moving in multiple degrees of freedom. The dynamic model including those features, along with contact dynamics between the robot and ground, is desirable for future implementation of control strategies and locomotion over complex terrain. Experiments are also performed on two centimeter-scale robot prototypes to validate the robot dynamics when using a simplified foot-terrain interaction model. The third task is to improve micro-robot performance using the resulting understanding of the robots’ nonlinear hybrid dynamics. One such step is the development of optimization algorithms for robot inputs. Acquisition of on-board sensing information of robot motion is first addressed. Then the performance of optimization algorithms with and without such information can help determine the importance of on-board sensing. The swarm performance of robots is also studied using swarm optimization, which is implemented in simulation using representative on-board sensing signals. Another application is to develop alternate potential locomotion gaits for a single robot from a design standpoint. Both applications are studied using rapid prototyping and simulations, which are then extended to predictions for true micro-scale robots.