A flight control system for aerial robots: algorithms and experiments

This paper presents a hierarchical flight control system for unmanned aerial vehicles. The proposed system executes high-level mission objectives by progressively substantiating them into machine-level commands. The acquired information from various sensors is propagated back to the higher layers for reactive decision making. Each vehicle is connected via standardized wireless communication protocol for scalable multi-agent coordination. The proposed system has been successfully implemented on a number of small helicopters and validated in various applications. Results from waypoint navigation, a probabilistic pursuit-evasion game and vision-based target tracking demonstrate the potential of the proposed approach toward intelligent flying robots.     

Human Face Detection in Digital Video Using SVMEnsemble

This Letter proposes automatic human face detection in digital video using a support vector machine (SVM) ensemble to improve the detection performance. The SVM ensemble consists of several independently trained SVMs using randomly chosen training samples via a bootstrap technique. Next, they are aggregated in order to make a collective decision via a majority voting scheme. Experimental results show that the proposed face detection method using SVM ensemble outperforms conventional methods such as using only single SVM and Multi-Layer Perceptron in terms of classification accuracy, false alarms, and missing rates.     

Complete and Rapid Regrasp Planning with Look-up Table

A pick-and-place operation in a 3-dimensional environment is a basic operation for humans and multi-purpose manipulators. However, there may be a difficult problem for such manipulators. Especially, if the object cannot be moved with a single grasp, regrasping, which can be a time-consuming process, should be carried out. Regrasping, given initial and final poses of the target object, is a construction of sequential transition of object poses that are compatible with two poses in the point of grasp configuration. This paper presents a novel approach for solving the regrasp problem. The approach consists of a preprocessing and a planning stage. Preprocessing, which is done only once for a given robot, generates a look-up table which has information on kinematically feasible task space of the end-effector throughout the entire workspace. Then, using this table, the planning automatically determines a possible intermediate location, pose and regrasp sequence leading from the pick-up to put-down grasp. With a redundant robot, it is shown experimentally that the presented method is complete in the entire workspace and can be implemented in real-time applications due to rapid regrasp planning time. The regrasp planner was combined with an existing path.     

Burst Mode Bandwidth Allocation for Real-Time Messages in IEEE 802.12 Networks

In IEEE 802.12 network, a repeater arbitrates among transmission requests on a round robin basis. A node can transmit a packet only when it is granted by the repeater. Recently, the IEEE 802.12 Committee approved a burst mode in which a node can transmit multiple packets per grant. In burst mode a bandwidth, or the number of packets per grant should be allocated taking into account the timing constraints of real-time messages. This paper proposes bandwidth allocation algorithms based on the burst mode to guarantee the deadlines of periodic real-time messages. Once the bandwidth of each node is derived from the lengths and deadlines of periodic messages, a node is allowed to transmit packets up to the allocated bandwidth per grant. Experimental results show that the proposed algorithms providemuch higher guarantee ratio and graceful degradation for heavy network load, compared with the existing approaches.     

On the Optimality of Randomized Deadlock Avoidance Policies

This paper revisits the problem of selecting an optimal deadlock resolution strategy, when the selection criterion is the maximization of the system throughput, and the system is Markovian in terms of its timing and routing characteristics. This problem was recently addressed in some of our previous work, that (i) provided an analytical formulation for it, (ii) introduced the notion of randomized deadlock avoidance as a generalization of the more traditional approaches of deadlock prevention/avoidance, and detection and recovery, and (iii) provided a methodology for selecting the optimal randomized deadlock avoidance policy for a given resource allocation system (RAS) configuration. An issue that remained open in the problem treatment of that past work, was whether the proposed policy randomization is essential, i.e., whether there exist any RAS configurations for which a randomized deadlock avoidance policy is superior to any other policy that does not employ randomization. The work presented in this paper establishes that for the basic problem formulation where the only concern is the (unconstrained) maximization of the system throughput—or the other typical performance objectives of minimizing the system work-in-process and mean sojourn time—randomization of the deadlock resolution strategy is not essential. However, it is also shown that, sometimes, it can offer an effective mechanism for accommodating additional operational constraints, like the requirement for production according to a specified product mix. Furthermore, the undertaken analysis provides an analytical characterization of the dependence of the aforementioned performance measures on the transition rates relating to the various events of the underlying state space, which can be useful for the broader problem of synthesizing efficient scheduling policies for the considered class of resource allocation systems.     

Application of megasonic cleaner to sub-micron particles of SIL optical flying head

 As the demand for large capacity storage drives has increased, the flying height of near-field recording (NFR) sliders becomes as small as 100 nm for super high storage density. Accordingly contamination problems have been a concern of information storage industry because it may cause a serious damage to solid immersion lens (SIL) of optical flying head. Sub-micro contaminants in air bearing of the NFR slider may affect the flyability and stability of the optical slider. In addition, the cleaning of small particles becomes more difficult as the contaminant particle size decreases because the adhesion force increases very much as the particle size decreases. Recently developed high-frequency ultrasonic (megasonic) cleaning technique have made it possible to remove sub-micron particles less than 100 nm without surface erosion and many remarkable results have been reported. In this paper, the megasonic technique is applied and tested for the cleaning of the flying head of NFR drive. 1 MHz ultrasound with maximum 100 W was used to remove polystyrene latex (PSL) particles and alumina particles deposited on the surface of the slider. Effective cleaning performance was observed without any damages on the slider surface using optical microscope and AFM at different ultrasonic energy levels and cleaning times. Received: 20 June 2002 / Accepted: 9 September 2002     

Discontinuity-Preserving Surface Reconstruction Using Stochastic Differential Equations

We address the problem of reconstructing a surface from irregularly spaced sparse and noisy range data while concurrently identifying and preserving the significant discontinuities in depth. It is well known that, starting from either the probabilistic Markov random field model or the mechanical membrane or thin plate model for the surface, the solution of the reconstruction problem can be eventually reduced to the global minimization of a certain “energy” function. Requiring the preservation of depth discontinuities makes the energy function nonconvex and replete with multiple local minima. We present a new method for obtaining discontinuity-preserving reconstruction based on the numerical solution of an appropriate Ito vector stochastic differential equation (SDE). The reconstructed surface is found by following the sample path of the (stochastic) diffusion process that solves the SDE in question. Our central contribution is the demonstration of the efficacy of the stochastic differential equation technique for solving a vision problem. Through comparisions of the results of our method to those of the two well-known existingglobalminimization based reconstruction techniques, we show a significant improvement in the final reconstructions obtained.     

Biped Locomotion by Reduced Ankle Power

Power reduction in the ankle joints of a biped robot is considered inthis paper. Ankles of human beings have small torque and are veryflexible within a certain range of motion (very stiff near and beyondthis range). This characteristic makes foot landing soft and gives agood contact between its sole and the ground. This feature can beimplemented in a biped robot by using a small actuator for the anklejoints. A small actuator consumes less energy and reduces the weightof the leg. With less power in the ankle joints, robot walkingbecomes more difficult to control. This problem can be solved byproviding a feedback control mechanism as presented in this paper. Thecontrol mechanism uses the motion of the body and the swinging leg toeliminate instability caused by the weak ankle. Two locomotionexamples, standing and walking, were investigated respectively toshow the validity of the proposed control scheme. In standing, thecontrol input is the displacement of the ankle joint of thesupporting leg. The control mechanism decides the bending angle ofthe body and the position of the swinging leg. For walking, only thebending angle of the body is used to avoid the discontinuity of thecontrol input. Experimental results are presented to show theeffectiveness of the control mechanism.     

Online learning via congregational gradient descent

We propose and analyse a populational version of stepwise gradient descent suitable for a wide range of learning problems. The algorithm is motivated by genetic algorithms which update a population of solutions rather than just a single representative as is typical for gradient descent. This modification of traditional gradient descent (as used, for example, in the backpropogation algorithm) avoids getting trapped in local minima. We use an averaging analysis of the algorithm to relate its behaviour to an associated ordinary differential equation. We derive a result concerning how long one has to wait in order that, with a given high probability, the algorithm is within a certain neighbourhood of the global minimum. We also analyse the effect of different population sizes. An example is presented which corroborates our theory very well.This work was supported by the Australian Research Council.     

Efficient inventory management in multi-echelon distribution systems

In this paper, an improved DRP method to schedule multi-echelon distribution network is proposed such that order-quantities and order points are dynamically obtained to meet the demand in just in time concept and minimize the out-of-stock probability. The order scheduling method reflects the dynamic characteristics of inventory level changes in the regional distribution centers and the central distribution center. The experiment has been done with various demand distributions, forecast error distributions and lead times. The proposed method was compared with the traditional DRP-based scheduling methods which use different lot-sizing and order point decision techniques. From the result, it is found that the proposed heuristic method yields preferable lot-sizing schedules.