Kourosh Parsa

Autonomous capture of a tumbling satellite

In this paper, we describe a framework for the autonomous capture and servicing of satellites. The work is based on laboratory experiments that illustrate the autonomy and remote-operation aspects. The satellite-capture problem is representative of most on-orbit robotic manipulation tasks where the environment is known and structured, but it is dynamic since the satellite to be captured is in free flight. Bandwidth limitations and communication dropouts dominate the quality of the communication link. The satellite-servicing scenario is implemented on a robotic test-bed in laboratory settings. The communication aspects were validated in transatlantic tests.

Design and Implementation of a Mechatronic, All-Accelerometer Inertial Measurement Unit

This paper discusses the design, calibration, simulation, and experimental validation of a kinematically redundant inertial measurement unit that is based solely on accelerometers. The sensor unit comprises 12 accelerometers, two on each face of a cube. The location and direction of the sensors are determined so as to locally optimize the numerical conditioning of the system of governing kinematic equations. The orientational installation error of each sensor is identified by off-line iterative processing of the gravitational acceleration measurements made at a number of known orientations of the unit, thus allowing subsequent calibration. Furthermore, a novel procedure is developed through which the acceleration measurements can be used to directly determine the body angular velocity; this results in a major accuracy improvement over similar works whereby the angular velocity is obtained via integrating the angular acceleration. Experimental results are presented to validate the methodology, design, and implementation.

A Reconfigurable Robot With Lockable Cylindrical Joints

This paper presents a new conceptual design for reconfigurable robots. Unlike conventional reconfigurable robots, our design does not achieve reconfigurability by utilizing modular joints. Rather, the robot is equipped with passive joints, i.e., joints without actuator or sensor, which permit changing the Denavit-Hartenberg (DH) parameters such as the link length and twist angle. The passive joints will become controllable when the robot forms a closed kinematic chain. Also, each passive joint is equipped with a built-in brake mechanism that is normally locked, but the lock can be released whenever the parameters are to be changed. Such a versatile and agile robot is particularly suitable for space application for its simple, compact, and light design. The kinematics and recalibration of this kind of reconfigurable robot are thoroughly analyzed. A stable reconfiguration-control algorithm is devised to take the robot from one configuration to another by directly regulating the passive joints to the associated, desired DH parameters. Conditions for the observability and the controllability of the passive joints are also derived in detail.

Motion and Parameter Estimation of Space Objects Using Laser-Vision Data

free-falling tumbling satellite (target). The filter receives only noisy pose measurements from a laser-vision system aboard another satellite (chaser) at a close distance in a neighboring orbit. The filter estimates the full states, all the inertia parameters of the target satellite, and the covariance of the measurement noise. A comprehensive dynamics model that includes aspects of orbital mechanics is incorporated for accurate estimation. The discrete-time model, whichinvolvesastate-transitionmatrixandthecovarianceofprocessnoise,isderivedinclosedform,thusrendering the filtersuitableforreal-timeimplementation.Thestatisticalcharacteristicsofthemeasurementnoiseisformulated by a state-dependent covariance matrix. This model allows additive quaternion noise, while preserving the unitnorm property of the quaternion. The convergence properties of the developed filter is demonstrated by simulation andexperimental results. These results also demonstrate that the filter can continuously produce accurate estimates of pose even when the vision system is occluded for tens of seconds.

An adaptive vision system for guidance of a robotic manipulator to capture a tumbling satellite with unknown dynamics

This paper is focused on an adaptive vision system for the guidance of a robot to intercept a non-cooperative target satellite with unknown dynamics parameters. A Kalman filter is developed to reliably estimate the states of the object as well as all of its inertial parameters - namely, the moment-of-inertia ratios, the center-of-mass location, and the orientation of the principle axes - from vision information. The estimates are then used to optimally plan the motion of the manipulator. The optimization performance index includes the time of travel and the weighted norms of the end-effector velocity and acceleration, and it is subject to the conditions that the robot end-effector and the satellite gasping point arrive at the rendezvous point with the same velocity and that the interception occurs within the robot reach. The variational method is used to find the optimal path, which turns out to be the solution of a fourth-order differential equation. Subsequently, a closed-form solution is obtained. The solution to the optimal terminal-time problem is also obtained from the Hamiltonian of the entire system. Experiments are conducted by using a robotic arm to move a satellite mockup according to orbital mechanics and measuring the satellite pose by a laser camera system. The results demonstrate a successful grasping even though the inertial parameters are not known by the control system.

Attitude calibration of an accelerometer array

An accelerometer-array attitude-calibration method is proposed. Being based on the compatibility of rigid-body point accelerations, this method can be applied to determine and account for the accelerometer installation errors with a high degree of accuracy. It is assumed that the number or three-axis accelerometers in the array is redundant in order to help reduce the effect of sensor noise, thereby obviating the Kalman-filtering of the signals. Procedures are developed to calculate the angular velocity and acceleration as well as the attitude of the body, all in the body frame. It is demonstrated that even large attitude errors can be dealt with via off-line iterative applications of the scheme.

Pose-and-twist estimation of a rigid body using accelerometers

An algorithm for estimating the pose and twist of a rigid body using measurements made by a redundant number of on-board accelerometers is formulated. Redundancy helps reduce the effect of the accelerometer noise. This algorithm is based only on rigid-body kinematics because it is intended to be used for rigid bodies acted upon by forces and moments elusive to modeling. Procedures are developed to calculate the angular velocity and acceleration as well as the attitude of the body along with the velocity and position of the centroid of the pickup points, all in the body frame. It is shown that a small installation error can cause the integration results to become unstable. This instability is compensated for by means of a pose-measurement sensor.

Adaptive motion estimation of a tumbling satellite using laser-vision data with unknown noise characteristics

A noise-adaptive variant of the Kalman filter is presented for the motion estimation and prediction of a free-falling tumbling satellite as seen from a satellite in a neighboring orbit. A complete dynamics model, including aspects of orbital mechanics, is incorporated for accurate estimation. Moreover, a discrete-time model of the entire system which includes the state-transition matrix and the covariance of process noise are derived effectively in a closed form, which is essential for the real-time implementation of the Kalman filter. We will show that the translational and rotational measurements are coupled and consequently derive the corresponding observation matrix. The statistical characteristics of the measurement noise is formulated by a state-dependent covariance matrix. This model allows additive quaternion noise, while preserving the unit-norm property of the quaternion. The estimator takes the noisy measurements from a laser vision system with unknown and possibly varying statistical noise properties, and subsequently the estimator adaptively estimates the full sates, i.e., the pose and the velocities, in addition to the covariance of the measurement noise and the inertial parameters of the target satellite. Simulations and experiments conducted will demonstrate the quality performance of the adaptive estimator.

Design and mechatronic implementation of an accelerometer-based, kinematically redundant inertial measurement unit

This paper discusses the design, calibration, and simulation of a kinematically redundant inertial measurement unit which is based solely on accelerometers. The sensor unit comprises twelve accelerometers, two on each face of a cube. The location and direction of the sensors are determined so as to locally optimize the numerical conditioning of the system of governing kinematic equations. The orientational installation error of each sensor is identified by off-line iterative processing of the gravitational-acceleration measurements made at a number of known orientations of the unit. Furthermore, a novel procedure is developed through which the acceleration measurements can be used to directly determine the body angular velocity; this results in a major accuracy improvement over similar works whereby the angular velocity is obtained via integrating the angular acceleration. Numerical results are presented

Configuration Control and Recalibration of a New Reconfigurable Robot

The advantages of reconfigurable robots have been discussed in the specialized literature. Conventionally, reconfigurability was a direct result of using modular joints. In this paper we discuss the configuration control and recalibration of a different class of reconfigurable robots, one which is equipped with lockable cylindrical joints with no actuators or sensors. Such a robot can be as versatile and agile as a hyper-redundant manipulator, but with a simpler, more compact, lighter design. A passive joint becomes controllable when the robot forms a closed kinematic chain and the joint lock is released. After reconfiguration, the values of the passive joints are computed from the value of the active joints using inverse kinematics of the closed chain. That problem is solved using a globally uniformly asymptotically stable scheme based on closed-loop inverse kinematics (CLIK). An asymptotically stable reconfiguration controller is also devised that takes the robot from one configuration to another by directly regulating the values of the passive joints. The controller has a rather simple structure, which only relies on the robot gravity and kinematics models. Conditions for the observability and the controllability of the passive joints are also derived in detail, and some numerical results are reported.