Volume 14, Number 5, 2017
Special Issue on Human-inspired Computing (pp.501-575)
The paper reviews and extends an emerging body of theoretical results on deep learning including the conditions under which it can be exponentially better than shallow learning. A class of deep convolutional networks represent an important special case of these conditions, though weight sharing is not the main reason for their exponential advantage. Implications of a few key theorems are discussed, together with new results, open problems and conjectures.
To achieve the artificial general intelligence (AGI), imitate the intelligence? or imitate the brain? This is the question! Most artificial intelligence (AI) approaches set the understanding of the intelligence principle as their premise. This may be correct to implement specific intelligence such as computing, symbolic logic, or what the AlphaGo could do. However, this is not correct for AGI, because to understand the principle of the brain intelligence is one of the most difficult challenges for our human beings. It is not wise to set such a question as the premise of the AGI mission. To achieve AGI, a practical approach is to build the so-called neurocomputer, which could be trained to produce autonomous intelligence and AGI. A neurocomputer imitates the biological neural network with neuromorphic devices which emulate the bio-neurons, synapses and other essential neural components. The neurocomputer could perceive the environment via sensors and interact with other entities via a physical body. The philosophy under the "new" approach, so-called as imitationalism in this paper, is the engineering methodology which has been practiced for thousands of years, and for many cases, such as the invention of the first airplane, succeeded. This paper compares the neurocomputer with the conventional computer. The major progress about neurocomputer is also reviewed.
The recently introduced theory of practopoiesis offers an account on how adaptive intelligent systems are organized. According to that theory, biological agents adapt at three levels of organization and this structure applies also to our brains. This is referred to as tri-traversal theory of the organization of mind or for short, a T3-structure. To implement a similar T3-organization in an artificially intelligent agent, it is necessary to have multiple policies, as usually used as a concept in the theory of reinforcement learning. These policies have to form a hierarchy. We define adaptive practopoietic systems in terms of hierarchy of policies and calculate whether the total variety of behavior required by real-life conditions of an adult human can be satisfactorily accounted for by a traditional approach to artificial intelligence based on T2-agents, or whether a T3-agent is needed instead. We conclude that the complexity of real life can be dealt with appropriately only by a T3-agent. This means that the current approaches to artificial intelligence, such as deep architectures of neural networks, will not suffice with fixed network architectures. Rather, they will need to be equipped with intelligent mechanisms that rapidly alter the architectures of those networks.
The center of mass (CoM) is a key descriptor in the understanding and the analysis of bipedal locomotion. Some approaches are based on the premise that humans minimize the CoM vertical displacement. Other approaches express walking dynamics through the inverted pendulum model. Such approaches are contradictory in that they lead to two conflicting patterns to express the CoM motion:straight line segments for the first approaches and arcs of a circle for the second ones. In this paper, we show that CoM motion is a trade-off between both patterns. Specifically, CoM follows a "curtate cycloid", which is the curve described by a point rigidly attached to a wheel rolling on a flat surface. We demonstrate that all the three parameters defining a curtate cycloid only depend on the height of the subjects.
Grasping is a significant yet challenging task for the robots. In this paper, the grasping problem for a class of dexterous robotic hands is investigated based on the novel concept of constrained region in environment, which is inspired by the grasping operations of the human beings. More precisely, constrained region in environment is formed by the environment, which integrates a bio-inspired co-sensing framework. By utilizing the concept of constrained region in environment, the grasping by robots can be effectively accomplished with relatively low-precision sensors. For the grasping of dexterous robotic hands, the attractive region in environment is first established by model primitives in the configuration space to generate offline grasping planning. Then, online dynamic adjustment is implemented by integrating the visual sensory and force sensory information, such that the uncertainty can be further eliminated and certain compliance can be obtained. In the end, an experimental example of BarrettHand is provided to show the effectiveness of our proposed grasping strategy based on constrained region in environment.
This paper describes a brain-inspired simultaneous localization and mapping (SLAM) system using oriented features from accelerated segment test and rotated binary robust independent elementary (ORB) features of RGB (red, green, blue) sensor for a mobile robot. The core SLAM system, dubbed RatSLAM, can construct a cognitive map using information of raw odometry and visual scenes in the path traveled. Different from existing RatSLAM system which only uses a simple vector to represent features of visual image, in this paper, we employ an efficient and very fast descriptor method, called ORB, to extract features from RGB images. Experiments show that these features are suitable to recognize the sequences of familiar visual scenes. Thus, while loop closure errors are detected, the descriptive features will help to modify the pose estimation by driving loop closure and localization in a map correction algorithm. Efficiency and robustness of our method are also demonstrated by comparing with different visual processing algorithms.
Inspired by the movement of duck that is able to move on land and water utilizing its webbed feet, a novel design of an amphibious robot has been presented in this paper. In contrary, the orthodox design of amphibious robot utilizes the tracks or wheels on land and switches to the propeller to move in water. The proposed design employs same propulsion system as webbed feet to move on land and water. After studying the movement of the duck underwater, a conclusion has been drawn that it is swimming in the water by moving its webbed feet back and forth to generate force to push its body forward. Recreating this phenomenon of duck movement, hybrid robot locomotion has been designed and developed which is able to walk, swim and climb steps using the same propulsion system. Moreover, webbed feet would be able to walk efficiently on muddy, icy or sandy terrain due to uneven distribution of robot weight on the feet. To be able to justify the feasibility of the design, simulations are being carried out using SimulationXpress of the SOLIDWORKS software.
In this paper, an efficient technique for optimal design of digital infinite impulse response (IIR) filter with minimum passband error (ep), minimum stopband error (es), high stopband attenuation (As), and also free from limit cycle effect is proposed using cuckoo search (CS) algorithm. In the proposed method, error function, which is multi-model and non-differentiable in the heuristic surface, is constructed as the mean squared difference between the designed and desired response in frequency domain, and is optimized using CS algorithm. Computational efficiency of the proposed technique for exploration in search space is examined, and during exploration, stability of filter is maintained by considering lattice representation of the denominator polynomials, which requires less computational complexity as well as it improves the exploration ability in search space for designing higher filter taps. A comparative study of the proposed method with other algorithms is made, and the obtained results show that 90% reduction in errors is achieved using the proposed method. However, computational complexity in term of CPU time is increased as compared to other existing algorithms.
This paper describes the evaluation method of the gait motion in walk rehabilitation. We assume that the evaluation consists of the classification of the measured data and the prediction of the feature of the gait motion. The method may enable a doctor and a physical therapist to recognize the condition of the patients more easily, and increase the motivation of patient further for rehabilitation. However, it is difficult to divide the gait motion into discrete categories, since the gait motion continuously changes and does not have the clear boundaries. Therefore, the self-organizing map (SOM) that is able to arrange the continuous data on the almost continuous map is employed in order to classify them. And, the feature of the gait motion is predicted by the classification. In this study, we adopt the gravity-center fluctuation (GCF) on the sole as the measured data. First, it is shown that the pattern of the GCF that is obtained by our developed measurement system includes the feature of the gait motion. Secondly, the relation between the pattern of the GCF and the feature of the gait motion that the doctor and the physical therapist evaluate by visual inspection is considered using the SOM. Next, we describe the prediction of following features measured by numerical values:the length of stride, the velocity of walk and the difference of steps that are important for the doctor and the physical therapist to make a diagnosis of the condition of the gait motion in walk rehabilitation. Finally, it is investigated that the position of a new test data that is arranged on the map accords with the prediction. As a consequence, we confirm that the method using the SOM is often useful to classify and predict the condition of the patient.
The stability analysis and anti-windup design problem is investigated for two linear switched systems with saturating actuators by using the single Lyapunov function approach. Our purpose is to design a switching law and the anti-windup compensation gains such that the maximizing estimation of the domain of attraction is obtained for the closed-loop system in the presence of saturation. Firstly, some sufficient conditions of asymptotic stability are obtained under given anti-windup compensation gains based on the single Lyapunov function method. Then, the anti-windup compensation gains as design variables are presented by solving a convex optimization problem with linear matrix inequality (LMI) constraints. Two numerical examples are given to show the effectiveness of the proposed method.
Through the direct parameter approach, a solution for spacecraft attitude tracking is presented. First of all, the spacecraft attitude tracking control model is built up by the error equation of the second-order nonlinear quaternion-based attitude system. Based on the control model, a suitable controller is designed by the direct parameter approach. Compared with other control strategies, the direct parameter approach can offer all degrees of freedom for the controller to satisfy the requirements for system properties and turn the original nonlinear system into closed-loop linear system. Furthermore, this paper optimizes the controller according to the robustness, limitation of controller output and closed-loop eigenvalue sensitivity. Putting the controller into the original system, the state response of the closed-loop system and the output of controller are plotted in Matlab to verify the availability and robustness of the controller. Therefore, the controlled spacecraft can achieve the goal of tracking on the mobile target with the external disturbance torque.