This paper will present an online approach for friction approximation to be utilized in con- cert with whole body control on humanoid robots. This approach allows humanoid robots with ankle mounted force-torque sensors to extrapolate information about the friction constraints at the hands during multi-contact poses without the addition of hardware to the platform. This is achieved by utilizing disturbance detection as a method of monitoring active forces at a single external point and deriving available friction force at said contact point in accordance with Coulomb's Law of Friction. First, the rigid body dynamics and required compliant humanoid model optimization are established which allow incorporation of friction constraints. These friction constraints are then informed by monitoring of external forces, which can be used as an indicator of slip based on tangential force. In practice, the robot with operational multi-contact whole body control is navigated to the desired contact surface and normal force only contact is initiated. Using an iterative coefficient estimation based on the achieved system forces, the robot tests the boundaries of its operable force range by inducing slip. Slip detection is utilized as the basis for coefficient estimation, which allows the robot to further understand its environment and apply appropriate forces to its contact points. This approach was implemented on a simple 3 link model to verify expected performance, and then on both the simulated model of Virginia Tech's ESCHER robot and in practice on the actual ESCHER platform. The proposed approach was able to achieve estimation of slip parameters, based largely on time spent measuring, actual friction coefficient, and the available contact force. Though the performance of the proposed approach is dependent on a number of variables, it was able to provide an operational parameter for the robot's whole body controller, allowing expansion of the support region without risking multi-contact slip. / Master of Science / This paper presents an approach for humanoid robots to use their hands to approximate the friction parameters of contact surfaces without prior knowledge of those parameters. This is accomplished as part of the robot’s control system and integrated into its balancing and movement operating system so that it may determine these parameters without ceasing operation. The proposed approach relies on the force sensors typically embedded in the ankles of bipedal robots as its sole force input, so no additional hardware need be added to the robot in order to employ this functionality. Once placed in contact, the robot is able to approximate the forces at its hand with these sensors, and use those approximate values as the basis for estimating the static friction coefficient of the system, in accordance with Coulomb’s Law of Friction. The robot’s onboard controller is able to utilize this information to ensure that it does not overestimate the available force that may be applied at the contact point, using prior knowledge of the robot model’s range of motion. In practice, the robot with this functionality is navigated to the desired contact surface and a hand contact that does not risk slip is initiated. Using an iterative coefficient estimation based on the achieved system forces, the robot tests the boundaries of its operable force range by inducing slip. Slip detection is utilized as the basis for coefficient estimation, which allows the robot to further understand its environment and apply appropriate forces to its contact points. This approach was implemented on a simple 3 link robot model to verify expected performance, and then on both the simulated model of Virginia Tech’s ESCHER robot and in practice on the actual ESCHER platform. The proposed approach was able to achieve estimation of slip parameters, based largely on time spent measuring, actual friction coefficient, and the available contact force. Though the performance of the proposed approach is dependent on a number of variables, it was able to provide an operational parameter for the robot’s whole body controller, allowing expansion of the support region without risking multi-contact slip.
| Identifer | oai:union.ndltd.org:VTETD/oai:vtechworks.lib.vt.edu:10919/78716 |
| Date | 18 August 2017 |
| Creators | Ridgewell, Cameron Patrick |
| Contributors | Mechanical Engineering, Lattimer, Brian Y., Furukawa, Tomonari, Tokekar, Pratap, Asbeck, Alan T. |
| Publisher | Virginia Tech |
| Source Sets | Virginia Tech Theses and Dissertation |
| Detected Language | English |
| Type | Thesis |
| Format | ETD, application/pdf |
| Rights | In Copyright, http://rightsstatements.org/vocab/InC/1.0/ |
Page generated in 0.0021 seconds