In this paper we consider the problem of estimating the state of a system with perspective outputs. We formulate the problem in a deterministic setting by searching for the value of the state that is “most compatible ” with the dynamics, in the sense that it requires the least amount of noise to explain the measured output. We show that, under appropriate observability assumptions, the optimal estimate converges globally to the true value of the state and can be used to design output-feedback controllers by using the estimated state to drive a state-feedback controller. We apply these results to the estimation of position and orientation of a controlled rigid body, using measurements from a charged-coupled-device camera attached to the body. 1

This paper addresses the state estimation of systems with perspective outputs. We derive a minimum-energy estimator which produces an estimate of the state that is “most compatible” with the dynamics, in the sense that it requires the least amount of noise energy to explain the measured outputs. Under suitable observability assumptions, the estimate converges globally asymptotically to the true value of the state in the absence of noise and disturbance. In the presence of noise, the estimate converges to a neighborhood of the true value of the state. These results are also extended to solve the estimation problem when the measured outputs are transmitted through a network. In that case, we assume that the measurements arrive at discrete-time instants, are time-delayed, noisy, and may not be complete. We show that the re-designed minimum-energy estimator preserves the same convergence properties. We apply these results to the estimation of position and orientation for a mobile robot that uses a monocular charged-coupled-device (CCD) camera mounted onboard to observe the apparent motion of stationary points. In the context of our application, the estimator can deal directly with the usual problems associated with vision systems such as noise, latency and intermittency of observations. Experimental results are presented and discussed.

This paper addresses the state estimation of continuous-time systems with perspective outputs, whose measurements arrive at discrete-time instants, are time-delayed, noisy, and may not be complete. Resorting to dynamic programming, we derive a minimum-energy estimator which produces an estimate of the state that is “most compatible” with the dynamics, in the sense that it requires the least amount of noise energy to explain the measured outputs. The state-estimator has the desired property that, under suitable observability assumptions, the estimate converges asymptotically to the true value of the state in the absence of noise and disturbance. In the presence of noise, the estimate remains bounded away from the true value of the state. We apply these results to the estimation of position and orientation for a mobile robot that uses a monocular chargedcoupled-device (CCD) camera mounted on-board to observe the apparent motion of stationary points. In the context of our application, the estimator can deal directly with the usual problems associated with vision systems such as noise, latency and intermittency of observations. Experimental results are presented and discussed.

This paper addresses the problem of nonlinear filter design to estimate the relative position and attitude of an autonomous vehicle with respect to a desired coordinate system defined by visual landmarks using measurements from an inertial measurement unit (IMU) and a monocular charged-coupled-device (CCD) camera mounted on-board. We formulate the problem in the framework of state estimation of a state-affine system with implicit outputs. Resorting to dynamic programming, we derive a H∞ estimator which produces an estimate of the state that is compatible with the dynamics and ensures a prescribed bound γ on the discounted induced L2-gain from disturbances and noise to estimation error. In our formulation we take directly into account that the measurements arrive at discrete-time instants, are time-delayed, and may not be complete. In this way, we can deal with usual problems in vision systems such as noise as well as latency and intermittency of observations. The convergence of the proposed observer system is analyzed and simulations results are presented and discussed.