N. Cuperlier, P. Laroque, P. Gaussier, and M. Quoy (France)
Motor control, navigation and mapping, animat approach, planning, bio-inspired system.
This paper reports on a comparative study on planning us ing a cognitive map. We propose an interesting alternative to model relying on cognitive map linking place cells. Our experiments show that using transitions is at least as sim ple and less costly (expensive) as using place cells, and that the number of transitions created tends to be stable over time. Then we propose a single mechanism based on Neural Fields for merging the different proposed transitions into a single effective transition used to plan. The work is illustrated by simulations underlining properties already used on real robot control.
Important Links:
Go Back
