Jingyan Song

Complex recurrent neural network for computing the inverse and pseudo-inverse of the complex matrix

A complex recurrent neural network (CRNN) is formulated and applied to compute the complex matrix inverse in real time. Both full rank and rand deficient matrices are considered. This paper extends recent works which apply real recurrent networks for real-valued matrix inversion.

A new hybrid navigation algorithm for mobile robots in environments with incomplete knowledge

Focusing on the navigation problem of mobile robots in environments with incomplete knowledge, a new hybrid navigation algorithm is proposed. The novel system architecture in the proposed algorithm is the main contribution of this paper. Unlike most existing hybrid navigation systems whose deliberative layers usually play the dominant role while the reactive layers are only simple executors, a more independent reactive layer that can guarantee convergence without the assistance of a deliberative layer is pursued in the proposed architecture, which brings two benefits. First, the burden of the deliberative layer is released, which is beneficial to guaranteeing real-time property and decreasing resource requirement. Second, some possible layer conflicts in the traditional architecture can be resolved, which improves the system stability. The convergence of the new algorithm has been proved. The simulation results show that compared with three traditional algorithms based on different architectures, the new hybrid navigation algorithm proposed in this paper performs more reliable in terms of escaping from traps, resolving conflicts between layers and decreasing the computational time for avoiding time out of the control cycle. The experiments on a real robot further verify the validity and applicability of the new algorithm.

Performance evaluation and optimization of human control strategy

Abstract Modeling human control strategy (HCS) is becoming an increasingly popular paradigm in a number of different research areas, ranging from robotics and intelligent vehicle highway systems to expert training and virtual reality computer games. Usually, HCS models are derived empirically, rather than analytically, from real-time human input–output data. While these empirical models offer an effective means of transferring intelligent behaviors from humans to robots and other machines, there is a great need to develop adequate performance criteria for these models. It is our goal in this paper to develop several such criteria for the task of human driving. We first collect driving data from different individuals through a real-time graphic driving simulator that we have developed, and identify each individual’s control strategy model through the flexible cascade neural network learning architecture. We then define performance measures for evaluating two aspects of the resultant HCS models. The first is based on event analysis, while the second is based on inherent analysis. Using the proposed performance criteria, we demonstrate the procedure for evaluating the relative skill of different HCS models. Finally, we propose an iterative algorithm for optimizing an initially stable HCS model with respect to independent, user-specified performance criteria, by applying the simultaneously perturbed stochastic approximation (SPSA) algorithm. The methods proposed herein offer a means for modeling and transferring HCS in response to real-time inputs, and improving the intelligent behaviors of artificial machines.

Adaptive Sliding Mode Fault-Tolerant Control of the Uncertain Stewart Platform Based on Offline Multibody Dynamics

In this paper, we propose a novel adaptive sliding mode fault-tolerant control scheme based on offline multibody dynamics for the uncertain Stewart platform under loss of actuator effectiveness. The asymptotic stability is analyzed by Lyapunov method in the presence of friction, unmodeled dynamics, environmental disturbances, and even the unpredictable actuator faults. To cope with the nonlinear coupling and various properties of freedom directions, the offline nominal multibody dynamics are employed to design the initial upper bound of uncertainties and to realize the dynamic compensation, which achieves high online computational efficiency and significantly improves the characteristics of the six degree-of-freedom (DOF) directions. We also introduce a novel adaptive updating law to adjust the control torque based on the real-time position tracking errors, which alleviates the chattering phenomenon of the sliding mode controller. Finally, the fault-free and faulty conditions are analyzed to corroborate the advantages of the proposed control scheme in comparison with the nominal sliding mode control scheme.

Real‐time motion planning for mobile robots by means of artificial potential field method in unknown environment

– The purpose of this paper is to focus on the local minima issue encountered in motion planning by the artificial potential field (APF) method, investigate the currently existing approaches and analyze four types of previous methods. Based on the conclusions of analysis, this paper presents an improved wall‐following approach for real‐time application in mobile robots., – In the proposed method, new switching conditions among various behaviors are reasonably designed in order to guarantee the reliability and the generality of the method. In addition, path memory is incorporated in this method to enhance the robots cognition capability to the environment. Therefore, the new method greatly weakens the blindness of decision making of robot and it is very helpful to select appropriate behaviors facing to the changeable situation. Comparing with the previous methods which are normally considering specific obstacles, the effectiveness of this proposed method for the environment with convex polygon‐shaped obstacles has been theoretically proved. The simulation and experimental results further demonstrate that the proposed method is adaptable for the environment with convex polygon‐shaped obstacles or non‐convex polygon‐shaped obstacles. It has more widely generality and adaptiveness than other existed methods in complicated unknown environment., – The proposed method can effectively realize real time motion planning with high reliability and generality. The cognition capability of mobile robot to the environment can be improved in order to adapt to the changeable situation. The proposed method can be suitable to more complex unknown environment. It is more applicable for actual environment comparing with other traditional APF methods., – This paper has widely investigated the currently existed approaches and analyzes deeply on four types of traditional APF methods adopted for real time motion planning in unknown environment with simulation works. Based on the conclusions of analysis, this paper presents an improved wall‐following approach. The proposed method can realize real time motion planning considering more complex environment with high reliability and generality. The simulation and experimental results further demonstrate that the proposed method is adaptable for the environment with convex polygon‐shaped obstacles or non‐convex polygon‐shaped obstacles. It has more widely generality and adaptiveness than other existed methods in complicated unknown environment.

An applicable short-term traffic flow forecasting method based on chaotic theory

Short-term traffic flow forecasting plays a very important role in urban traffic management and control. In this paper, According to the chaotic property of urban traffic flow, we compute the parameters of phrase space reconstruction for traffic flow system. Meanwhile, a local-forecasting method is introduced to predict urban road short-term traffic flow based on the theory of phrase space reconstruction. Self-organizing Map (SOM) network is introduced to seek the near neighbor. Case study using real traffic flow data from UTC-SCOOT system proves the validity of the method. The research in this paper is a significant attempt to forecast traffic flow from the viewpoint of non-linear time series.

Topology optimization for urban traffic sensor network

Abstract This paper presents an optimized topology for urban traffic sensor networks. Small world theory is used to improve the performance of the wireless communication system with a heterogeneous transmission model and an optimal transmission radius. Furthermore, a series of simulations based on the actual road network around the 2nd Ring Road in Beijing demonstrate the practicability of constructing artificial “small worlds”. Moreover, the particle swarm optimization method is used to calculate the globally best distribution of the nodes with the large radius. The methods proposed in this paper will be helpful to the sensor nodes deployment of the new urban traffic sensor networks.

Dynamic modeling of a 6-degree-of-freedom Stewart platform driven by a permanent magnet synchronous motor

For an electrical six-degree-of-freedom Stewart platform, it is difficult to compute the equivalent inertia of each motor in real time, as the inertia is time-varying. In this study, an analysis using Kane’s equation is undertaken of the driven torque of the movements of motor systems (including motor friction, movements of motor systems along with the actuators, rotation around axis of rotors and snails), as well as driven torque of the platform and actuators. The electromagnetic torque was calculated according to vector-controlled permanent magnet synchronous motor (PMSM) dynamics. By equalizing the driven torque and electromagnetic torque, a model was established. This method, taking into consideration the influence of counter electromotive force (EMF) and motor friction, could be applied to the real-time dynamic control of the platform, through which the calculation of the time-varying equivalent inertia is avoided. Finally, simulations with typically desired trajectory inputs are presented and the performance of the Stewart platform is determined. With this approach, the multi-body dynamics of the electrical Stewart platform is better understood.

Three-tiered sensor networks architecture for traffic information monitoring and processing

With the advent and applications of sensor networks, pervasive traffic information can be gathered rapidly and collaboratively, which can upgrade traditional traffic monitoring system and enable many unprecedented services for travelers. In this paper, three-tiered sensor network architecture is proposed to approach the novel traffic information service system that would dramatically improve the ways and qualities of traffic information collection. A brief analysis of the important requirements of system, as well as the key issues that are faced in the design process, is described. The detailed strategies at each level, from the network architecture, to the sensor unit configuration and software deployment, to the operation of system, are explained. The implementation of a pilot platform is discussed in the end.

A new method for mobile robots to avoid collision with moving obstacle

A new method for helping mobile robots to avoid collisions with moving obstacles is proposed. This method adopts the concept of safe sectors in the vector field histogram (VFH) method, but simplifies its description. Moreover, the new method takes the threat of moving obstacles into account when selecting the direction of motion, and a new speed control law that considers more factors is designed. Hence, it is better at avoiding moving obstacles than the VFH method. Simulation results indicate that the new method also shows many advantages over the dynamic potential field (DPF) method, which is a representative approach for avoiding moving obstacles. Experiments have further verified its applicability.