Abstract: We explore how different types of memory can aid spatial navigation in
changing uncertain environments. In our simple foraging task, every
day, the agent has to find its way from its home, through barriers, to
food. The world is non-stationary: from day to day, the location of
some barriers or food may change. The agent’s sensing is limited, and
its location information is uncertain. Any map construction, and use
such as planning, needs to be robust against such uncertainties. Any
learning should be adequately fast. We look at a range of strategies,
from simple to sophisticated, with various uses of memory. We find
that the agent that builds and keeps updating a map, even though the
map is partial and noisy, can be substantially more efficient than the
simpler agents, as task difficulties such as distance to goal are
raised, as long as the uncertainty, from localization and change, is
not too large.
Loading