SSP 1995 project summary: | |
Un-manned missions using controlled or semi-autonomous robots have been proposed as a more practical and less expensive alternative to manned exploration of the moon and Mars. Such robots tend to be large and expensive requiring complex telemetary and hierarchical descision-making algorithms. An alternative approach which has been suggested is to use a large number of small, cheap, entirely autonomous robots which employ simple proximity sensors and descision rules to achieve the same mission goals.
A particularly appropriate application of these would be to prospect and mine minerals or water on the lunar surface. A typical scenario would be to land a mother vehicle carrying a large number of these robots near the southern lunar pole. The robots would then random walk around the region looking for water. On locating a deposit, a robot would send a transponder signal indicating the presence of water to the other robots which would then move in the direction of the signal. Each robot would collect up a load of water and return to the mother craft.
Simulations using up to 100 robots with a simple terrain topology and descision making algorithms have been carried out by Dr C.R Mcinnes. This project has involved developing and implementing a more complex serial implementation of the model involving a realistic terrain and more complex robot behaviour. A real-time visualisation tool using a Silicon Graphics workstation was then developed. Finally, a parallel version version of the implementation was implemented targetted at a cluster of workstations.
Owen O'Loan worked on this project.
Compressed PostScript of the project's final report is available here (821341 bytes) .
Webpage maintained by mario@epcc.ed.ac.uk |