A compliant force/torque sensor for robot force control has been developed. This thesis presents methods of designing, testing, and implementing the sensor on a robotic system. The sensor uses an orthoplanar spring equipped with Hall-effect sensors to measure one component of force and two moment components. Its unique design allows for simple and cost effective manufacturing, high reliability, and compactness. The device may be used in applications where a robot must control contact forces with its environment, such as in surface cleaning tasks, manipulating doors, and removing threaded fasteners. The compliant design of the sensor improves force control performance and reduces impact forces. Sensor design considerations are discussed, followed by a discussion of the proposed design concept. Theoretical compliance and stress analysis of the orthoplanar spring is presented that allows for rapid design calculations; these calculations are validated via finite element analysis. A mechanical design method is given which uses the results of the compliance and stress analysis. Transducer design is then addressed by developing a model of the sensor. The design methods are used to design a prototype sensor which is tested to determine its instrument uncertainty. Finally, the sensor is implemented on a robotic platform to test its performance in force control.

In recent years, increased emphasis has been placed on the development of parallel-architecture mechanisms for use as robotic manipulators. Parallel robots offer the benefits of higher load-carrying capacity, greater positioning accuracy and lower weight when compared to serial devices. However, robotic wrist development has traditionally focused on serial mechanisms having a large, spherical workspace and simpler kinematic solutions. The Carpal wrist is a unique parallel mechanism consisting of a fixed base and a movable output plane connected via three serial kinematic chains. The forward and inverse kinematic problems of the Carpal wrist are solved closed-form, making the device suitable for use as a new type of robotic wrist. The closed-form solutions are dependent upon the assumptions that the fixed and moving planes are symmetric about a mid-plane and that the three kinematic chains connecting the planes are identical. This thesis investigates the errors that result from those assumptions being violated due to manufacturing and assembly errors. In the non-ideal model, pose error is found by iteratively solving a system of equations describing the output plane position and orientation and comparing them with the ideal solution. The error model is a tool for predicting the effects of kinematic parameter errors on the positioning accuracy and reachable workspace of the Carpal wrist. In this work, a general error model is developed and validated for a range of parameter error values. Special-case results are presented for errors in the individual parameters. / Master of Science