A hierarchical neurocontroller architecture, which comprises two artificial neural network (ANN) systems for inversion kinematics solution and motion control of robotic redundant manipulators is presented. The solution of inverse kinematics is realized by a Hopfield network, in which the global planning of a collision-free trajectory is based on the potential functions using the necessary conditions of minimum for an integral type criterion. A direct servo-level controller is utilized by a multilayer feedforward network based on backpropagation algorithm, in which the computed torque technique is employed to control manipulator's joints to track the trajectory. The stability of the both sub-controllers is analyzed. Another major contribution of this paper is to provide an approach for the most difficult problem in using neurocontroller-how to efficiently train the designed ANNs.

A mathematical model for an adaptive expert system in anesthesia is presented. The concept of clusters that utilize clinical attributes in order to reduce the dimensionality of the patient's state-space is introduced. One goal of the model is to implement the existing categories and to identify and cluster categories as well as make the system adaptive to new and more optimal categories. Well-known techniques of pattern classification and cluster analysis are used on the measurable dataset to look for new categories or to readjust existing ones. Readjustment is required to optimize the existing categories to give the most efficient classification of the diseases.>

The vital function status (VFS), a real-time indicator that continuously depicts the overall status of the patient under anesthetic, is presented. The VFS is generated by a subsystem which has been developed as a part of adaptive real-time expert system for anesthesia. The subsystem can operate independently and may be implemented in other highly monitored environments. It presents data at higher level of abstraction and does not hinder the accessibility to routine data (the routine monitors are still in use). On the contrary, more data per unit of time is being considered. >

Summary form only given. Previous work where estimation of the parameters of a plant was incorporated through exploratory schedules (ESs) which are reference input trajectories designed to enhance the learning of system parameters, is extended. In that work, ESs were generated offline and used in an open-loop fashion. Moreover, these ESs were used in between actual control tasks, therefore limiting the process of estimation during idle time. In this work the authors present an approach for generating ESs in a closed-loop manner. Such trajectories in general may not be the desired trajectories, since they result in larger tracking errors. However, ESs offer faster convergence to the system parameters and therefore yield smaller long-term tracking errors. The automation for the design of ESs requires online modification of the desired trajectory to enhance learning at the expense of poorer initial tracking.>

A nonalgorithmic method is presented for the solution to the inverse kinematic problem of a robot. The method is robot independent and involves a hybrid approach whereby a neural solution is augmented with an iterative procedure which provides the final solution within some specified tolerance. Essentially, the neural solution is similar to a lookup table in providing a good initial guess to a classical iterative search. It has been found that for the industrial manipulator PUMA 560, the proposed hybrid method achieves about a twofold increase in computational efficiency with better uniformity of the time required to obtain the solution to the robotic manipulator.>

The slack set technique, an algorithm for determining a minimum-time obstacle avoidance trajectory for a robot in a known environment, is presented. For time-optimal trajectories with constrained acceleration and velocity, the shortest time of motion may be different for each joint or axis of the system. Thus some delay of a joint other than the slowest will not necessarily affect the time of motion for the entire system. This natural redundancy for obstacle avoidance is exploited in order to simplify the trajectory search algorithm by at least one order of magnitude. By neglecting the presence of all obstacles and assigning maximum control to each actuator (bang-bang control), a lower-bound estimate of the time needed to complete a task (T/sub task/) is calculated. The A* heuristic search is used to search what is named the slack set, namely, a subset of the state space that contains only those states that are members of a trajectory with a task time equal to T/sub task/. If no trajectory is found during the initial search, the subset of the state-space being examined is sequentially increased until a valid trajectory is found. The slack set technique is guaranteed to find a feasible monotonic trajectory if such a trajectory exists in the slack set. Since, in general, the minimum-time obstacle avoidance trajectory is not unique, secondary constraints such as minimum distance, minimum distance in the state space, and others can also be satisfied. The method is demonstrated via a planar mobile robot.>

An important class of neural models is described as a set of coupled nonlinear differential equations with state variables corresponding to the axon hillock potential of neurons. Through a nonlinear transformation, these models can be converted to an equivalent system of differential equations whose state variables correspond to firing rates. The firing rate formulation has certain computational advantages over the potential formulation of the model. The computational and storage burdens per cycle in simulations are reduced, and the resulting equations become quasilinear in a large significant subset of the state space. Moreover, the dynamic range of the state space is bounded, alleviating the numerical stability problems in network simulation. These advantages are demonstrated through an example, using the authors' model for the so-called neural solution to the traveling salesman problem proposed by J.J. Hopfield and D.W. Tank (1985).>

An architecture for an adaptive neuromorphic system designed to control a robot is suggested. The proposed architecture utilizes two important features of neural networks: the abundance of local minima in the network's state space and the uniformity of convergence of these minima in the face of growing dimensionality. The proposed approach is expected to yield controllers which are both faster and simpler than controllers which are designed by the methods of model reference adaptive control and self-tuning regulator. The controller's complexity is expected not to grow exponentially with the number of unknown parameters, and to allow adaptation in both continuous and discrete parameter domains. The possible benefits of the architecture are demonstrated on a single-degree-of-freedom manipulator, whose controller is assisted by a neural estimator.>

An algorithm for determining the minimum-time obstacle avoidance trajectory for a robot is presented. Since different joints, when moved independently, may reach their desired end values at different times, any delay of a joint, other than the slowest, will not affect the total time of motion. This natural redundancy is used with obstacle avoidance to simplify any path search algorithm by at least one order of magnitude (one degree of freedom less). By neglecting the presence of all obstacles and assigning to each actuator maximum control torque (bang-bang), a lower-bound estimate of the time T/sub task/ needed to complete a task is calculated. The A/sup */ heuristic search is used to search the subset of the state space which contains only those states which are members of a trajectory with a task time equal to T/sub task/. If no trajectory is found during the initial search, the subset of the state space being examined is sequentially increased until a valid trajectory is found. Since, in general, the minimum-time obstacle avoidance trajectory is not unique, secondary constraints such as minimum distance in the state space and others can also be satisfied.>

The stability, capacity, and design of a nonlinear continuous neural network are analyzed. Sufficient conditions for existence and asymptotic stability of the network's equilibria are reduced to a set of piecewise-linear inequality relations that can be solved by a feedforward binary network, or by methods such as Fourier elimination. The stability and capacity of the network is characterized by the postsynaptic firing rate function. An N-neuron network with sigmoidal firing function is shown to have up to 3/sup N/ equilibrium points. This offers a higher capacity than the (0.1-0.2)N obtained in the binary Hopfield network. It is shown that by a proper selection of the postsynaptic firing rate function, one can significantly extend the capacity storage of the network.>

A new algorithm of minimum time motion planning is proposed for robots operating in the obstacle strewn environment. Structure of the topological passageways is analyzed and represented using a model of slalom situations for which a number of rules is determined. Dynamical system of robot is described in a form of sequential machine. This enabled a merger between two kindred algorithms: A* search algorithm, and dynamic programming. An experimental analysis of the simulated mobile robot has confirmed the applicability of results.