Traditional coarse pointing, acquisition, and tracking (CPAT) systems are pre-calibrated to have the center pixel
of the camera aligned to the laser pointing vector and the center pixel is manually moved to the target of interest
to complete the alignment process. Such a system has previously demonstrated its capability in aligning with
distant targets and the pointing accuracy is on the order of sensor resolution. However, aligning with targets at
medium range where the distance between angular sensor and transceiver is not negligible is its Achilles Heel.
This limitation can be resolved by imposing constraints, such as the trifocal tensor (TT), which is deduced from
the geometrical dependence between cameras and transceivers.
Two autonomous CPAT systems are introduced for FSO transceiver alignment in mid- and long-range scenarios.
This work focuses on experimental results that validate the pointing performance for targets at different
distances, backed up by the theoretical derivations. A mid-range CPAT system, applying a trifocal tensor as its
geometric invariant, includes two perspective cameras as sensors to perceive target distances. The long-range
CPAT system, applying linear mapping as the invariant, requires only one camera to determine the pointing
angle. Calibration procedures for both systems are robust to measurement noise and the resulting system can
autonomously point to a target of interest with a high accuracy, which is also on the order of sensor resolution.
The results of this work are not only beneficial to the design of CPAT systems for FSO transceiver alignment,
but also in new applications such as surveillance and navigation.
Using imaging sensors in a pointing, acquisition and tracking (PAT) system provides a high degree of accuracy and the capability of controlling multiple transceivers simultaneously. However, this kind of system can suffer from a narrow field-of-view (FOV). Also, for a single image sensor, the resulting PAT system can only accurately acquire a target located far away or moving on a plane. In this paper, we describe an optical PAT system, which has a FOV of almost 180 degrees and is also capable of locating targets at any distance. It is well known that two regular cameras are sufficient to reconstruct three dimensional coordinates by triangulating with two incident rays. The uniqueness of our system is that one of the cameras is an omnidirectional fisheye camera and the other is a regular camera (with a FOV of 30 degrees). Their geometric relationship is encoded by a radial trifocal tensor, which is further discussed in this paper. This scheme leads to a hierarchical structure in the design of PAT systems. A regular camera and an omnidirectional camera form a camera pair. Each of the transceivers on a rotary gimbal and the camera pair is first calibrated in order to retrieve their individual radial trifocal tensor. After the calibration, the camera pair monitors the target of interest and further computes the rotation angles for each of transceivers. A transceiver is then selected and rotates toward the target. The selection process is based on network objectives. The resulting system is fully optical and has three distinct advantages: 1) the radial trifocal tensor is invariant with the motion of the entire platform, 2) a wide field-of-view (close to 180 degrees); 3) three dimensional acquisition capability. However, there is a small penalty to be paid because an additional geometric limit exists compared with a PAT system using two regular cameras.
A free space optical (FSO) network consists of many reconfigurable, directional, high data-rate links. Its performance can be optimized by using topology control algorithms, which involve: (1) potential neighbor information collection, (2) an optimization algorithm with given constraints, and (3) a precise pointing procedure. In general, if a sensor at each node can observe a large field of view (FOV), then more potential link targets can be detected. With more possible link choices, the optimization algorithm will have greater degrees of freedom in determining the optimum topology. The intuitive way to acquire a wide spatial acquisition range is to use a camera with a wide FOV. However, for such a wide angle lens/mirror, there are inevitable large aberrations, which cause errors in a pointing procedure based on image analysis. To mitigate these aberrations, a possible solution is to build a correction procedure from the wide FOV lens imaging model to a pinhole imaging model. In this context, a mapping model is proposed, based on analyses of several wide angle lens sets using CodeV. The proposed model also compensates for the effect of deviations between the center lines of the lens and a CCD imaging array. To obtain the optimum parameters of the model, an off-line calibration procedure based on geometrical constraints is introduced. A sensor system consisting of a widely available fisheye converter (Nikon FC-E8) and a high-resolution CCD camera (1392x1040 pixels) has been built for evaluating the model's performance, as part of our pointing, acquisition and tracking (PAT) system.
A free space optical (FSO) link requires precise pointing, acquisition and tracking (PAT) for reliable communication. Among the various tracking schemes, optical tracking with an imaging system offers great precision. For a point-to-point FSO link, the tracking problem has six degree of freedom (DOF), including position coordinates and orientation angles. By using imaging systems at both end of the link this can be reduced to a problem with three DOF. By converting from Cartesian to spherical coordinates, we can further reduce the problem to two DOFs. Thus, a single camera at each end of the link is sufficient. In this paper, we propose an outside-in, real-time tracking system, which is based on blob extraction and two-dimensional prediction. The system is inspired by the stereo vision algorithm from the computer vision community. It can be divided into two parts: (1) the tracker, and (2) the pointer. They are operated in a closed-loop, which stabilizes performance and accuracy. The tracker is used for extracting target information. Its accuracy has to be in the milliradian range, which provides a first constraint. To operate in real-time, an optical beacon is placed at each end of the link, which is imaged as a "blob". The size of this "blob" imposes a second constraint on the tracker. To work within these constraints, a 1.4Mpixel digital camera is used. The pointing system consists of two
stepping motors with resolution of 0.125 mrads and slew speed up to 800,000 steps per second. The overall system covers a whole sphere in less then 1 second.
Hybrid free space optical/radio frequency (FSO/RF) networks promise broadband connectivity, high availability and quality of service (QoS), together with the capability of autonomous reconfigurability to deal with changing atmospheric and traffic conditions in dynamic environments. Nodes with n-connectedness (multiple transceivers) offer great flexibility in constructing new network topologies. Moreover, topologies using hybrid links are more effective in changing atmospheric conditions than those, using either communication modality alone. While FSO links can be expected to be available >99% of the time on links up to 1km in length, high performance RF provides backup connectivity in heavily obscured conditions. We have designed and implemented gimbal-mounted, hybrid FSO/RF nodes with combined apertures for joint pointing, acquisition, and tracking (PAT) operation. These nodes incorporate directional RF antennas for PAT network setup and management, and FSO links for very high data rate transmission. We describe these hybrid nodes and their performance, our hybrid network simulations, and our re-configurable network testbed for high data rate video transmission. Our simulations include realistic modeling of obscuration, traffic management, and topology control to deal with link non-availability and optimization of network performance. Hybrid, directional networks are scalable and provide low probability of intercept/detection (LPI/LPD) operation, especially in FSO mode.
Free space, dynamic, optical wireless communications will require topology control for optimization of network performance. Such networks may need to be configured for bi- or multiple-connectedness, reliability and quality-of-service. Topology control involves the introduction of new links and/or nodes into the network to achieve such performance objectives through autonomous reconfiguration as well as precise pointing, acquisition, tracking, and steering of laser beams. Reconfiguration may be required because of link degradation resulting from obscuration or node loss. As a result, the optical transceivers may need to be re-directed to new or existing nodes within the network and tracked on moving nodes. The redirection of transceivers may require operation over a whole sphere, so that small-angle beam steering techniques cannot be applied. In this context, we are studying the performance of optical wireless links using lightweight, bi-static transceivers mounted on high-performance stepping motor driven stages. These motors provide an angular resolution of 0.00072 degree at up to 80,000 steps per second. This paper focuses on the performance characteristics of these agile transceivers for pointing, acquisition, and tracking (PAT), including the influence of acceleration/deceleration time, motor angular speed, and angular re-adjustment, on latency and packet loss in small free space optical (FSO) wireless test networks.
The worldwide demand for broadband communications is being met in many places through the use of installed single-mode fiber networks. However, there is still a significant 'first-mile' problem, which seriously limits the availability of broadband Internet access. Free-space optical wireless communication has emerged as a technique of choice for bridging gaps in the existing high data rate communication networks, and as a backbone for rapidly deployable mobile wireless communication infrastructure. Because free space laser communication links can be easily and rapidly redirected, optical wireless networks can be autonomously reconfigured in a multiple-connected topology to provide improved network performance. In this paper we describe research designed to improve the performance of such networks. Using topology control algorithms, we have demonstrated that multiply-connected, rapidly reconfigurable optical wireless networks can provide robust performance, and a high quality of service at high data rates (up to and beyond 1 Gbps). These systems are also very cost-effective. We have designed and tested on the University of Maryland campus a prototype four-node optical wireless network, where each node could be connected to the others via steerable optical wireless links. The design and performance of this network and the topology control is discussed.