Robot path planning using swarms of active particles
There exist a wide range of techniques that generate collision-free (optimal) trajectories, in an autonomous fashion, for mobile robotics. Probably, the most popular technique is that of artificial potential fields, where the robot is treated as a particle subject to a potential field that is generated by the obstacles and the goal position. The path generation problem is then treated as an optimization problem where gradient descent methods have been traditionally used. Particle swarm optimization has also been widely used to solve optimization problems, and although it has proven to be more efficient in the search of minima, it also suffers from early convergence, i.e. the swarm may get trapped in a local minima. Many of the modifications that cater for this weakness involve added complexity in the construction of the potential field and thus translates into ad-hoc algorithms. A particle swarm approach with the property of escaping local minima by forcing vortex-like dynamics when the gradient of the potential field is close to zero is proposed, proving to be more compact and intuitive than previously proposed algorithms.
Robot path planning using swarms of active particles, Page 1 of 2
< Previous page Next page > /docserver/preview/fulltext/books/ce/pbce119f/PBCE119F_ch8-1.gif /docserver/preview/fulltext/books/ce/pbce119f/PBCE119F_ch8-2.gif