LOLA won the Office Cleanup event at the 1995 Robot Competition and Ex
hibition, held as part of the Fourteenth International Conference on A
rtificial Intelligence. The event called for a robot to pick up trash
in an unstructured environment and sort it such that the recyclable tr
ash winded up in the recycle bin and the regular trash in the trash bi
n. The only allowable information lola was given beforehand were model
-based descriptions of the trash and recyclables, which it located usi
ng color vision. Much of LOLA's success can be attributed to the simpl
e, fast algorithms and methods that also model sensor uncertainty. The
ideas and design philosophy that went into LOLA borrow heavily from t
hose of previous competitors-to which we are greatly indebted. These m
ethods and ideas are discussed here.