Search

We present a novel lipreading system that improves on the task of speaker-independent word recognition by decoupling motion and content dynamics. We achieve this by implementing a deep learning architecture that uses two distinct pipelines to process motion and content and subsequently merges them, implementing an end-to-end trainable system that performs fusion of independently learned representations. We obtain a average relative word accuracy improvement of ≈6.8% on unseen speakers and of ≈3.3% on known speakers, with respect to a baseline which uses a standard architecture.

Crowd behavior analysis plays an important role in high security interests in public areas such as railway stations, shopping centres, and airports, where large populations gather. The crowded scenes vary in various densities, structures and occlusion. It brings enormous challenges in identifying generic descriptors to describe motion dynamics caused by pedestrians walk in different directions with extremely diverse behaviors. Therefore, this research is proposal an approach for crowd behavior analysis to recognize the common properties across different crowded scenes. The recognized common properties are then used to identify generic descriptors from group-level for crowd behavior classification.

A system of packing bulky materials into fabric containers was developed, including a gripping mechanism, a motion drive, and a control system thereof, which allows performing any manipulatory operations with fabric containers where a bulky material is automatically filled in. The system design is very simple and implies a possible application of flexible containers of standard types and sizes. A trial model of the automatic packing system developed has been made. The procedure of the experimental research was developed to determine the efficiency, reliabity, and operation modes of the system. Experimental research of the motion dynamics of the gripping mechanism, which is the major unit of the automatic packing system, were carried out, its utility and working capacity were verified.

The subject and at the same time the main purpose of this article is to create a mathematical model describing the phenomenon of dynamic movement of the holder (the subject), used to support the visual unit (camera) fitted to the UAV platform. In addition, the Kalman filter used in the considerations improved the waveforms of control signals and reduced the influence of undesired signals on the control processes of other systems that are equipped with unmanned aircraft. Based on the analysis of the UAV object, the mathematical model created and selected simulation tests, the final part of the paper summarizes and formulates practical conclusions reflecting the actual work of the UAV object.

Indoor positioning solutions attempt to extend location-based services to indoor environments where typical outdoor positioning solutions are not applicable. Current mainstream indoor positioning solutions rely on extra devices such as WiFi APs and iBeacons. These solutions involve a non-trivial cost of installing and managing. To avoid such non-trivial cost, infrastructure-free geomagnetic based techniques have been proposed, however, only practically verified in limited indoor environments. In this study, we present our commercially adopted end-to-end geomagnetic sensing based solution including a cost-effective measurement methodology to collect geomagnetic field signals, an online calibration technique to address sensor distortion and diversity issues, and an online motion identification technique to resolve issues of motion dynamics. According to the results carried out in 120 shopping malls, our solution’s error distance is persistently within 5m.

In order to obtain an electrical model compatible with practical situations, modeling the load of electrical engines is fundamental. Hence, the motion dynamics related to railway traction must be considered. Some particular aspects are extremely important for a reliable model of the load, including track slope, resistance in curves and air drag force and mechanical friction. With an accurate model for the load, it is possible to estimate the electromagnetic torque to be generated by the locomotive traction system and consequently to properly specify the electrical drive systems.

In this paper, we present a problem of regulating the motion of a rolling ball in a one-dimensional space in the presence of non-linear effects of friction and contact. The regulation problem is solved using a model-based reinforcement learning technique. A Gaussian process model is learned to make predictions on the motion of the ball and then, the predictive model is used to solve for the control policy using dynamic programming by estimating the value functions. Several results are shown to demonstrate the simple, yet interesting motion dynamics for the ball. Our hope is that the proposed system will serve as a simple benchmark system for reinforcement and robot learning.

This paper addresses issues in fall detection from videos. Since it has been a broadly accepted intuition that a falling person usually undergoes large physical movement and displacement in a short time interval, the study is thus focused on measuring the intensity and temporal variation of pose change and body motion. The main novelties of this paper include: (a) characterizing pose/motion dynamics based on centroid velocity, head-to-centroid distance, histogram of oriented gradients and optical flow; (b) extracting compact features based on the mean and variance of pose/motion dynamics; (c) detecting human by combining depth information and background mixture models. Experiments have been conducted on an RGB-D video dataset for fall detection. Tests and evaluations show the effectiveness of the proposed method.

In this paper, we address the problem of recognizing human actions with motion dynamics alone. For this purpose, we propose to use silhouette sequences to represent the human actions by discarding the appearance information, and then model the sequences with linear dynamical systems (LDSs). Recognition is achieved by directly comparing the distance between LDSs, rather than resorting to complex Bayesian learning and inference. In particular, we introduce an efficient optimization method to learn robust LDSs, and develop a shift invariant distance metric to measure the similarity on the LDSs space. We evaluate our approach on the human action data set and achieve comparable results.

The problem of modeling and control for autonomous electric vehicles is considered. Mutual interactions among vehicle motion dynamics are evaluated. Drivability control is also considered for a hybrid electric vehicle and a pure electric vehicle to optimize the operating freedom.

Recently, haptics has begun to impact consumer products, e.g., mobile phones and automobiles. In this paper, we introduce one such new application, that is, haptic simulation of refrigerator operation, and present an initial prototype for the front door. A one degree-of-freedom haptic interface is designed and built to provide torque feedback of the front door. Simulation software consisting of system control, graphic renderer, and haptic renderer are also developed. For haptic rendering, the motion dynamics of a refrigerator door is modeled, and the haptic renderer is implemented based on the dynamics model. Lastly, we report a user experiment carried out to assess the perceived similarity between simulated and real door operations, i.e., the realism, which shows promising results.

We propose a new motion dynamics approach to solve the PnP problem, where a dynamic simulation system is constituted by springs and balls. The equivalence between minimizing the energy of the dynamic system and solving the PnP problem is proved. With the assumption of the existence of resistances, the solution of the original PnP problem can be solved through the simulation of the process of the movement of the balls.

The present paper proposes a new-generation route optimization technique for an oceangoing vessel. Although many studies for route optimization have been conducted, most have implicitly considered ship dynamics. The recent advances of computing performance, however, enable us to calculate ship motion dynamics. In the proposed method, numerical simulation is used to compute the objective function necessary for a route optimization model as a minimum time problem. To examine the robustness of the method, calculation results on a container linear on a trans-Pacific route as a test case are demonstrated. It is concluded that by using a Bezier curve, only a few control variables are needed to generate the optimum route. This technique is now formulated as a minimum time problem, but it could be applicable for various extended problems.

The problem of navigation and control for autonomous vehicles is considered. Mutual interactions among vehicle motion dynamics are evaluated. It is proposed the mathematical model suitable for describing and simulating the whole motion of autonomous passenger vehicles.

The problem of navigation and control for autonomous vehicles is considered. Mutual interactions among vehicle motion dynamics are evaluated. It is proposed the mathematical model suitable for describing and simulating the whole motion of autonomous passenger vehicles.

In this paper, nonlinear observer is designed to estimate vehicle motion dynamics and tire forces in both of the longitudinal and lateral direction. Steering wheel angle and individual angular wheel speed is measured and the other vehicle states are not measured. The proposed observer assures bounded estimation error. Observer design methodology is simple to implement and to be computed on a real-time basis. This observer may be proposed to enhance redundancy of the closed vehicle dynamics controller system and fault-tolerance capability of detecting potentially critical maneuver situations. Simulation results are presented to illustrate the effectiveness of the presented nonlinear observer.

In this paper we present a novel nonlinear observer structure for estimating the tire-road friction force for tire pressure monitoring by use of the friction parameters identification results of LuGre tire model based on ant colony algorithm (ACA) with high precision. Simulation results show its feasibility for the simplified motion dynamics of a quarter-vehicle model and it may offer a new method for the research of tire pressure monitoring.

The problem of modelling and control for autonomous vehicles is considered. Mutual interactions among vehicle motion dynamics are evaluated. It is proposed the mathematical model suitable for describing and simulating the whole motion of autonomous passenger vehicles.

In light of the high nonlinearity of LuGre friction model, a novel method based on ant colony algorithm (ACA) for identifying the friction parameters of LuGre tire model is proposed. ACA is a parallelized bionic optimization algorithm inspired from the behavior of real ants, and a kind of positive feedback mechanism is adopted in ACA. On the basis of brief introduction of LuGre friction model, a method for identifying the static LuGre friction parameters and the dynamic LuGre friction parameters using ACA is derived. Finally, this new friction parameter identification scheme is applied to the simplified motion dynamics of a quarter-vehicle model with high precision. Simulation and application results verify the feasibility and the effectiveness of the scheme. It provides a new way to identify the friction parameters of LuGre tire model.

The kinematically decoupled joint space decomposition is proposed as a general coordinate transformation between the joint space and the task space of kinematically redundant manipulators. The method can describe the kinematic, force, and dynamic redundancy in a unified way using a minimal number of systematic motion variables. Based on the method, a set of equivalent dynamical models of redundant manipulators is constructed and its characteristics are studied focusing on the dynamical decoupling between the task and the null motion dynamics.