| Overview of the proposed navigation system for autonomous mobile robots. For detailed description, refer to the overview page. |
| Dynamic Bayesian networks exhibit temporal continuity and can modify their structure during execution. For more details, please go to the Task 2 page. | |
| Beliefs are temporally propagated in dynamic Bayesian networks using a short term memory structure and a feedback loop. The conditional probablitity matrix on the feedback loop is used to weight the effect of the previous beliefs on the current belief. For more details, please go to the Task 2 page. |
| Beliefs about objects can also be propagated spatially through the occupancy grid. This is especially true if the object spans multiple grid cells or is moving. For more details, please go to the Task 3 page. | |
| We use dynamic occupancy grids as our map representation. This occupancy grid is overlayed with a potential map to facilitate the control algorithm. For detailed description, please see the Task 3 page. |
| We use potential field based control for low level navigation of the mobile robot. This has been implemented in simulation along with dead reckoning. We have performed experiments on navigating the wheelchair using potential gradient maps through static worlds. For more details, please go to the Task 4 page. | |
This page is maintained by Amit Singhal.
Last update: 11/19/97.