Escaping Depressions in LRTS Based on Incremental Refinement of Encoded Quad-Trees

Table 1

Performance measures of the algorithms to specify their amount of planning, state revisitations, heuristic updates, and runtime. These indicate whether EQ LRTS effectively speeds up heuristic learning.

Metric name	Semantics
NRVN	The number ratio of visited nodes of LRTS on quad-trees and uniform cells.; Smaller NRVN means that EQ LRTS possesses a more effective planning process and visits fewer nodes.
TRRV	The times ratio of repeated visits of LRTS on quad-trees and uniform cells, which is the core index to illuminate how many times the agents visit the nodes they have already traversed. Smaller TRRV means less dramatic “thrashing” and vice versa.
RUT	The ratio of updates times, which is an inner and microcosmic index to point out how many times the program has done the heuristic updates. This index has a similar effect to TRRV but is dependent on the heuristic function to some extent.
RR	The runtime ratio, which can be seen as a bottom line. Although more calculation and operations are involved in quad-tree modeling than uniform cells, running EQ LRTS is not expected to consume a lot more time than its counterpart.