next up previous contents
Next: Physically Based Shape Up: Active Perception Previous: Model Building and

Dynamically Estimating Pose from Invariant Feature Measurements

Complex robotic tasks such as autonomous exploration and grasping naturally entail the co-operation of multiple sensors and actuators. In order to integrate sensor measurements and actuator control schemes it is generally useful to know the rigid body transformations relating the native co-ordinate frames of the various subsystems. In this research we examine the problem of estimating pose of a robot-mounted range-finding camera. We use range measurements of several classes of invariant scene features and known robot kinematics to provide an estimate of camera pose relative to the robot end effector which is maximally consistent with the data. Accuracy and stability of the estimator are evaluated for simulated and real data sets. We have shown that this methodology is equally applicable to calibrating a multi-robot workcell, and have examined its relation to robot kinematic model calibration and visual feedback.

D. Baird, F.P. Ferrie

Thierry Baron
Mon Nov 13 10:43:02 EST 1995