Abstract Almost all entities—animate or inanimate—that we see around us change with time. The changes are brought about by changes in the values of their attributes. By using a set of parameters to represent the variable attributes of an entity, and by suitably manipulating their values at run time, the behaviour of an entity can be broadly mimicked in animation. The majority of entities, however, are all too complex to animate directly. They are better described in terms of nested layers of smaller and simpler entities, which we call components. Each component is structurally and behaviourally complete and can be described independent of its application. In the present paper, we propose a scheme for 3D animation that broadly follows this line. The keystone of this scheme is a language, nicknamed ‘V’, which defines the structural and visual attributes of each component of the scene and associates a parameterized behaviour with it, if necessary, in the form of a program script. Thereafter, wherever such a component appears, it does so with a built‐in behaviour, which can nevertheless be regulated by its higher‐level component through its parameters. The advantage is that an entire animation can be modelled in a declarative fashion in terms of nested components with embedded behaviour. Besides, each component is easy to write, alter and reuse. The effort for development, debugging and maintenance of animation modelled in this way is much less as the concerns are almost always local. Copyright © 2002 John Wiley & Sons, Ltd.

For mobile robot navigation in an unknown and changing environment, a reactive approach is both simple to implement and fast in response. A neural net can be trained to exhibit such a behaviour. The advantage is that, it relates the desired motion directly to the sensor inputs, obviating the need of modeling and planning. In this work, a feedforward neural net is trained to output reactive motion, in response to sonar range inputs, with data generated artificially on the computer screen. The authors develop input and output representations appropriate to this problem. A reactive robot, being totally insensitive to context, often gets trapped in oscillations in front of a wide object. To overcome this problem, the authors introduce a notion of memory into the network by including context units at the input layer. The authors discuss the mode of training for such a network and present simulated runs of the trained net in various situations. The authors conclude by demonstrating a short tour of an actual mobile robot in their laboratory under the control of the trained neural net.

A walking machine must plan its footsteps in such a way that it is able to follow the desired route without stepping over the bad patches on the ground. A good way of doing this is to search through a graph of possible moves. We have previously done that for a four-legged machine (1990, 1991). For a six-legged machine, however, there are too many ways of choosing a stable support pattern. So, strong heuristics are needed to reduce the search. In this paper, a novel heuristic rule, based on the support state transitions of the wave gait, has been presented. This rule helps in converging to and maintaining the wave gait through a limited search, while retaining adequate options to deviate from the wave gait, as and when needed, to adapt to the terrain, or to execute a generalised motion. Thus, we have a gait generation mechanism that combines the efficiency of the wave gait with the terrain adaptivity of the free gait.>

Redundant manipulators have more degrees of freedom than what is absolutely necessary for performing a task. The extra degrees of freedom can be used for avoiding obstacles or to optimize certain performance indices like manipulability or task compatibility. Maximizing manipulability keeps the manipulator away from singularities and provides more velocity transmission ratios in all directions. Optimizing task compatibility improves the force/velocity transmission ratios in the specified directions. However, the real time implementation of various optimizing algorithms is difficult because of the need of large computing time. In the present work, robot configurations for an optimum performance index are computed throughout the workspace. These configurations are then used to train a layered feed forward neural network (FFNN). During operation of the robot, the trained neural net outputs optimal configurations in real-time. The neural net captures the gross behaviour of the training data rather than memorizing the individual data, as in a lookup table. Thus its output is smooth and ideally suited for control purposes. We have simulated this approach on a 3-DOF redundant planar manipulator and the results are discussed in this paper.>

A method is presented for the generation of a free gait for the straight-line motion of a quadruped walking machine. It uses a heuristic graph search procedure based on the A* algorithm. The method essentially looks into the consequences of a move to a certain depth before actually committing to it. Deadlocks and inefficiencies are thus sensed well in advance and avoided.>

A method is presented for the generation of a locally optimal free gait for 2-D generalized motion of a quadruped walking machine. It employs a heuristic graph search procedure based on the A* algorithm. The method essentially looks into the consequences of a move to a certain depth before actually committing to it. Deadlocks and inefficiencies are thus sensed well in advance and avoided.>