| Advanced robot manipulators with autonomous capability to adapt to unforeseen system changes must do so in a timely way with minimal human intervention. In the robotics context, trajectory planning is a necessary procedure when obstacles are present. The procedure, however, becomes a complex and time-consuming venture in situations where the obstacles are geometrically complicated or the robot is kinematically redundant. A conglomerate-of-spheres method using potential fields was developed for dealing with these situations. The method promises to greatly reduce the computational time associated with calculating collision points, and thus makes real-time obstacle avoidance more achievable.;The method simplifies geometric complexity by using a small number of overlapping spheres to approximate a complex three-dimensional object, and thus accelerates the process of collision detection without requiring the elaborate geometric modeling of complex surfaces. A neural network mapping using the Restricted Coulomb Energy paradigm, with the incorporation of a merging algorithm, is employed to achieve this, such that an accurate geometric description of complex objects can be obtained with an efficient number of spheres. The potential field approach, in turn, takes the obstacles defined above and simplifies the kinematic redundancy by establishing a potential field with a spherical symmetry to each robot link and obstacle. Two types of potential are involved: the attractive potential consistently attracts the tip of the end-effector to its destination, while the repulsive potential repels any robot link within the spherical affecting range of an encountering obstacle. Feasible trajectories are then determined by static analyses and the Newtonian motion laws under the influence of the potential fields.;The method was implemented for a PUMA 560 arm manipulator. As a result, the method outperformed a benchmark alternative, the "solinterf" function in Autocad R12, with an average improvement of two orders of magnitude. Furthermore, the performance showed no deviation as the geometric complexity increased, while that of the benchmark degenerated dramatically. With the advances in sensor technology and the need to perform trajectory planning in an unknown environment, methods proposed with sensor-based interfaces for real-time adjustment of trajectories could be lucrative extensions of the current work into the future. |
