A reactive navigation system for an autonomous mobile robot in unstructured dynamic environments is presented. The motion of moving obstacles is estimated for robot motion planning and obstacle avoidance. A multisensor-based obstacle predictor is utilized to obtain obstacle-motion information. Sensory data from a CCD camera and multiple ultrasonic range finders are combined to predict obstacle positions at the next sampling instant. A neural network, which is trained off-line, provides the desired prediction on-line in real time. The predicted obstacle configuration is employed by the proposed virtual force based navigation method to prevent collision with moving obstacles. Simulation results are presented to verify the effectiveness of the proposed navigation system in an environment with multiple mobile robots or moving objects. This system was implemented and tested on an experimental mobile robot at our laboratory. Navigation results in real environment are presented and analyzed.

The problem of navigating a mobile robot among moving obstacles is usually solved on the condition of knowing the velocity of obstacles. However, it is difficult to provide such information to a robot in real time. In this paper, we present an environment predictor that provides an estimate of future environment configuration by fusing multisensor data in real time. The predictor is implemented by an artificial neural network (ANN) trained using a relative-error-backpropagation (REBP) algorithm. The REBP algorithm enables the ANN to provide output data with a minimum relative error, which is better than conventional backpropagation (BP) algorithms in this prediction application. The mobile robot can, therefore, respond to anticipated changes in the environment. The performance is verified by prediction simulation and navigation experiments.

In this paper we present a virtual force guidance (VFG) system for dynamic motion planning and navigation of a mobile robot. This new method is developed to work with a predicted environment, which is provided by an artificial neural network (ANN) using the information from on-board sensor system. The proposed ANN predictor is trained by a relative-error-backpropagation (REBP) algorithm derived in this paper. The REBP algorithm allows the outputs of an ANN to have minimum relative error, which is better than the conventional backpropagation algorithm in this particular application. The VFG system, which can react to the future environment, assumes that the goal attracts the robot and the future obstacles repulse it. The resultant force determines the desired change in motion. This motion is therefore dependent on both the current motion of the robot and the future environment. Both simulation and experimental results are presented to show our approach can effectively navigate the robot in a human-like fashion.

This paper proposes a simple, yet effective scheme to prevent congestion in a packet-switched multistage interconnection network (MIN) caused by hot spots. In this scheme, switches in the second and third stages of the MIN monitor their buffer occupancy to detect any notable nonuniform access behavior. When a switch detects congestion, packets generated by processors will be blocked from entering the congested switch until the congestion is cleared. Our scheme is compared with two well known schemes and shown to exhibit significantly better performance than these two.>

The beam-opening angle problem of an ultrasonic range sensor is solved by using artificial neural networks to fuse multiple ultrasonic sensors. A mathematical model about the same problem is also derived. Both simulation and experimental results of scanning images are presented. The line extractor of the sensor image is developed by using the regression model with a little modification. The simulation results show this method is useful for environment recognition of robotic systems.

A planning procedure is developed for the navigation of a mobile robot in an unknown environment. A modified Gilles algorithm is used for path planning based on the world model which can be constructed from or modified by sensory information. with this algorithm, the environment is modelled as cells generated by using the safe-margin of obstacles. this method results in faster search than the original Gilles algorithm. The quadratic method is used to search for the desired path in a region of quadrant that is determined by the start and goal positions. To construct an environment model by means of the sensory information, the method of using bit-mapping sequences is developed. It is further fit into the world model and used by the path planning procedure. Smooth motion planning is achieved by using the predictor method in order to facilitate a direct commanding to the real robot. A monitor is developed using the techniques of expert systems to run the computer simulation.>