In real world scenarios, guiding vision to focus on salient parts of the visual space is a computationally demanding tasks. Selective attention is a biologically inspired strategy to cope with this problem, that can be used in engineered systems with limited resources. In active vision systems however, the stringent realtime requirements limit the space of solutions that can be achieved with conventional machine vision techniques and systems. We propose a hybrid approach where we combine a custom neuromorphic VLSI saliency-map based attention system with a conventional imager and a workstation, to implement both fast contrast-based saccadic eye movements in parallel with standard machine vision attention models using high-resolution color input images. We describe the system and present its response properties using basic control experiments.
This paper presents findings from a descriptive research on social gaming. A video-enhanced diary method was used to understand the user experience in social gaming. From this experiment, we found that natural human behavior and gamer's decision making process can be elicited and speculated during human computer interaction. These are new information that we should consider as they can help us build better human computer interfaces and human robotic interfaces in future.
Many social robots in the forms of conversation agents or Chatbots have been put to practical use in recent years. Their typical roles are online help or acting as a cyber agent representing an organisation. However, there exists a new form of devious chatbots lurking in the Internet. It is effectively an interactive malware seeking to lure its prey not through vicious assault, but with seductive conversation. It talks to its prey through the same channel that is normally used for human-to-human communication. These devious chatbots are using social engineering to attack the uninformed and unprepared victims. This type of attacks is becoming more pervasive with the advent of Web 2.0. This survey paper presents results from a research on how this breed of devious Malware is spreading, and what could be done to stop it.
We present our contribution in projects related to ad-hoc networking using different routing protocols and hardware platforms, showing our results and new solutions regarding topology control and routing protocols. We mainly focus on our work in the GUARDIANS EU-project, where as a main disaster scenario a large industrial warehouse on fire is assumed. The paper presents the simulation results for the routing protocols ACR, DSR and EDSR, as well as the implementation of down-scaled demos for supporting the autonomous team of robots as well as the human squad team with robust communication coverage. Various hardware platforms were used in the demos for distance measurement, based on laser range finder and radio communication with time of flight analysis.
The industrial design cycle starts with design then simulation, prototyping, and testing. When the tests do not match the design requirements the design process is started over again. It is important for students to experience this process before they leave their academic institution. The high cost of the prototype phase, due to CNC/Rapid Prototype machine costs, makes hands on study of this process expensive for students and the academic institutions. This document shows that the commercially available LEGO NXT Robot kit is a viable low cost surrogate to the expensive industrial CNC/Rapid Prototype portion of the industrial design cycle.
This paper explores strategies for teaching robotics not simply as a subject in its own right, but, using robotics in the teaching environment as an opportunity to stimulate creative thinking and generating an interest in science and technology as creative endeavours. The spirit is very much that espoused by C.P.Snow in his attempts to bridge "the two cultures" i.e. that of the arts on the one hand and that of science and technology on the other.
Actuators and gear trains of most biped humanoid robots are divergently allocated on the links of two legs. Disadvantages of such a mechanical design are complicated wiring of power cord and sensing/ control signal bundles and imprecise kinetics models of mixed link-and-actuator structures. Based on these drawbacks, this paper proposes a tendon-driven mechanism to develop a lower body structure of a full-size biped humanoid robot. The actuators are compacted as an actuator module, and they are placed at a distal site. A 12 degree-of-freedom mechanical structure is proposed with 100 cm in height and 45 kg in weight. The gait planning module is simulated and evaluated using the Matlab software. At the same time, an ARM7 based controller is developed to automatically generate walking patterns as well as to control the motors. Finally, a tendon-driven biped humanoid robot prototype is realized for practical waling control in the future.
This paper proposes a multi-robot coordination architecture for dynamic task, role and behavior selections. The proposed architecture employs the motivation of task, the utility of role, a probabilistic behavior selection and a team strategy for efficient multi-robot coordination. Multiple robots in a team can coordinate with each other by selecting appropriate task, role and behavior in adversarial and dynamic environment. The effectiveness of the proposed architecture is demonstrated in dynamic environment robot soccer by carrying out computer simulation and real environment.
Synthetic skins with humanlike characteristic would make it possible to address some of the psychosocial requirements of prosthetic hands as well as the safety and acceptance issues in social robotics. This paper describes the development of three-dimensional finite element models of synthetic finger phalanges. With the aim of duplicating the skin compliance of human finger phalanges, the model was used to investigate the effects of (i) introducing open pockets in the internal structure and (ii) combining different materials as external and internal layers. The results show that having pockets in the internal structure of the design can increase the skin compliance of the synthetic phalanges and make it comparable with the human counterpart. Moreover, having different layers can be used to satisfy skin compliance and other design requirements such as wear and tear.
Automated human identification by their walking behavior is a challenge attracting much interest among machine vision researchers. However, practical systems for such identification remain to be developed. In this study, a machine learning approach to understand human behavior based on motion imagery was proposed as the basis for developing pedestrian safety information systems. At the front end, image and video processing was performed to separate foreground from background images. Shape-width was then analyzed using 2D discrete wavelet transformation to extract human motion features. Finally, an adaptive boosting (AdaBoost) algorithm was performed to classify human gender and age into its class. The results demonstrated capability of the proposed systems to classify gender and age highly accurately.
Face recognition is a very important aspect in developing human-robot interaction (HRI) for social robots. In this paper, an efficient face recognition algorithm is introduced for building intelligent robot vision system to recognize human faces. Dimension deduction algorithms locally linear embedding (LLE) and adaptive locally linear embedding (ALLE) and feature extraction algorithm scale-invariant feature transform (SIFT) are combined to form new methods called LLE-SIFT and ALLE-SIFT for finding compact and distinctive descriptors for face images. The new feature descriptors are demonstrated to have better performance in face recognition applications than standard SIFT descriptors, which shows that the proposed method is promising for developing robot vision system of face recognition.
It is interested for a bipedal robot manipulated by different Degree of Freedoms (DoFs) to spend how much motion energy. In this paper, the motion energy of a biped robot influenced by the manipulation of different DoFs is studied. For calculating the motion energy, the forward and inverse kinematics of the designed biped robot are first derived. The 4-3-4 trajectory for planning the movement of robot joints in smoothing is designed for the walking finished by the biped joint movement. Once the joint trajectories of the biped robot are solved, the motion energy including kinetic and potential energy can be calculated. The walking of the biped robot manipulated by 2, 4 and 6 DoFs, respectively, is also included. The study provides the mechanical design of biped robots in the future.
While working in a dynamic environment, humanoid robots are subject to unknown forces and disturbances, putting them at risk of falling down and damaging themselves. One mechanism by which humans avoid falling under similar conditions is the human momentum reflex. Although such systems have been devised, the processing requirements are too high to be implemented on small humanoids having microcontroller processing capabilities. This paper presents a simplified momentum controller for fall avoidence. The system is tested on a simulated robot developed under Gazebo as well as under a real humanoid. Results show successful fall avoidance.
This paper discusses the real-time optimal construction of DTM by two measures. One is to improve coordinate transformation of discrete points acquired from lidar, after processing a total number of 10000 data points, the formula calculation for transformation costs 0.810s, while the table look-up method for transformation costs 0.188s, indicating that the latter is superior to the former. The other one is to adjust the density of the point cloud acquired from lidar, the certain amount of the data points are used for 3D construction in proper proportion in order to meet different needs for 3D imaging, and ultimately increase efficiency of DTM construction while saving system resources.
Stereo and motion analysis are potential techniques for providing information for control or assistance systems in various robotics or driver assistance applications. This paper evaluates the performance of several stereo and motion algorithms over a long synthetic sequence (100 stereo pairs). Such an evaluation of low-level computer vision algorithms is necessary, as moving platforms are being used for image analysis in a wide area of applications. In this paper algorithms are evaluated with respect to robustness by modifying the test sequence with various types of realistic noise. The novelty of this paper is comparing top performing algorithms on a long sequence of images, taken from a moving platform.
In this paper we propose a method for vision only topological simultaneous localisation and mapping (SLAM). Our approach does not use motion or odometric information but a sequence of noisy visual measurements observed by traversing an environment. In particular, we address the perceptual aliasing problem which occurs using external observations only in topological navigation. We propose a Bayesian inference method to incrementally build a topological map by inferring spatial relations from the sequence of observations while simultaneously estimating the robot's location. The algorithm aims to build a small map which is consistent with local adjacency information extracted from the sequence measurements. Local adjacency information is incorporated to disambiguate places which otherwise would appear to be the same. Experiments in an indoor environment show that the proposed technique is capable of dealing with perceptual aliasing using visual observations only and successfully performs topological SLAM.
In this paper, we mainly investigate a fast algorithm, Extreme Learning Machine (ELM), on its equivalent relationship, approximation capability and real-time face detection application. Firstly, an equivalent relationship is presented for neural networks without orthonormalization (ELM) and orthonormal neural networks. Secondly, based on the equivalent relationship and the universal approximation of orthonormal neural networks, we successfully prove that neural networks with ELM have the property of universal approximation, and adjustable parameters of hidden neurons and orthonormal transformation are not necessary. Finally, based on the fast learning characteristic of ELM, we successfully combine ELM with AdaBoost algorithm of Viola-Jones in face detection applications such that the whole system not only retains a real-time learning speed, but also possesses high face detection accuracy.
In this paper, a novel algorithm for sensory area coverage by mobile robots is proposed, with applications in, for example, mapping and exploration. The algorithm generates a sequence of nodes for the robot to visit, such that it covers as much as possible of the arena using a laser range finder. A crucial part of the exploration behavior implementing this algorithm is a deadlock avoidance procedure, which allows the robot to handle the inevitable problems (such as coping with narrow passages and obstacles of different size and height) that occur when navigating in complex arenas. Despite the simplicity of the algorithm, the robot is generally able to cover (with its laser range finder) 98% or more of an initially unexplored arena.
It has been shown that a task-level controller with minimal-effort posture control produces human-like motion in simulation. This control approach is based on the dynamic model of a human skeletal system superimposed with realistic muscle like actuators whose effort is minimised. In practical application, there is often a degree of error between the dynamic model of a system used for controller derivation and the actual dynamics of the system. We present a practical application of the task-level control framework with simplified posture control in order to produce life-like and compliant reaching motions for a redundant task. The addition of a sliding mode controller improves performance of the physical robot by compensating for unknown parametric and dynamic disturbances without compromising the human-like posture.
We have been investigating depth estimation techniques using variant of depth of field (DOF) by tilted optics imaging. These techniques can be applied for robotic and automation tasks, because it needs extremely fewer multiple focus images for depth estimation than the conventional passive methods. Hence, our method gets rid of the bottleneck of passive methods with a single camera; the motion speed of optical mechanics is very slower than that of image processing parts. Therefore, it is suitable for the high speed depth estimation like real-time processing accomplished by hardware such as a smart image sensor or an FPGA.
