Force Guided Assembly Under Bias

Abstract

Abstract Humans are able to pump gas into a car with little or no difficulty. This task is characterized by two sources of force: that from the nozzle contacting the car and that from the hose attached to the pump. The task succeeds due to the appreciable skill of a human and a forgiveness in the connection. The robotic mating of connectors burdened by forces from sources like the gas hose is beyond the current state of art. The research presented in this dissertation develops technology for robots to mate connectors that concurrently experience appreciable forces from encumbrances, like those from hoses, cables and oscillating masses, in addition to forces from contact. Effective force guided assembly under the influence of these bias forces, requires the differentiation of contact forces from bias forces, a task that is impossible using traditional sensing configurations. Emulating the contribution of bias during contact allows the estimation of bias forces and, subsequently, contact forces. By measuring and modeling bias prior to contact, when the only forces on the connector are from bias, a model of the bias source can be made. This model can be used to emulate bias during contact, enabling the differentiation of contact forces and allowing gentle force guided assembly. This thesis asserts that identification of and compensation for biasing forces will enable the robotic assembly of complex and fragile connectors that would otherwise be impossible.