Geometrical, mechanical and transduction techniques for designing multi-axis force-torque sensors for robotic applications

Abstract

This project is part of the effort initiated by Vincent Duchaine, Professor at the laboratory of Control and Robotics (CoRo), to enhance the physical interaction between humans and robots. An important aspect of this initiative concerns the force feedback and the ability of the robot to feel the forces that are applied to it. To measure the forces that are exerted on, for instance, a humanoid robot’s wrist, multi-axis force-torque sensors are essential. With these sensors, the robot can detect obstacles, avoid collisions, and get the job safely done. We hypothesize that the ability to accurately detect the force measurements that are associated with force control will help to improve the dexterity of such robots. To this end, we aim to develop new methods for designing the compliant structures that form the core of multi-axis force-torque sensors. The resulting sensors must be highly sensitive to a sufficient range of forces, and they must be capable of measuring forces in all directions. Moreover, they must be compact, so that they can be easily integrated within robotic structures; and of course, they must be cost-effective. This thesis attempts to answer three questions: 1)What is the best performance index for evaluating the geometry of the compliant structure? 2)What is the best systematic method to use for designing the mechanical structure? and 3) What is the best transduction technique for achieving a compact sensor design? We began our research by investigating the geometric model of the sensor, to better understand the equations by which the force and torque are computed. Based on this analysis, we proposed two performance indices, which we then validated. We next investigated how the behavior and the mechanical proprieties of the structure can be used to convey accurate estimates of the forces. Through an advanced systematic design method, based on a symbolic formulation of the wrench-displacement relationship, we designed a compact and cost-effective triaxial force sensor. Our sensor uses a single capacitive sensing element to detect several forces efficiently. In order to improve the design of the sensing element, we proposed a new transduction technique. This technique allowed us to build a multi-axial sensing element that is able to separate stress components of normal stress, shear stress, and torque. This innovative use of a single multi-axial sensing element allowed us to reduce the size of the network of sensing elements, and consequently lower the total dimensions and cost of the multi-axis force-torque sensor. In a series of experiments, we verified that our multi-axial sensing element can successfully measure multiple forces, proving that it is ready for use in numerous robotic applications.

Date	14 Dec 2015
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Vincent Duchaine (Supervisor) & Philippe Cardou (Co-supervisor)