Unbiased Model-Agnostic Metalearning Algorithm for Learning Target-Driven Visual Navigation Policy

Table 1

Comparison results of standard deviation for SPL and SR in the same scene type.

Scenario	Method	Scene instance								SD
		bedroom01		bedroom02		bedroom03		bedroom04
		SPL	SR	SPL	SR	SPL	SR	SPL	SR	SPL	SR
Trained target in unseen environment	TDVG	8.56	13.7	0.81	4.11	0.56	3.89	9.41	15.7	4.81	6.23
	MPSL	6.87	10.8	11.4	19.9	2.51	8.44	3.41	11.2	4.03	5.02
	GCN	7.19	12.1	16.4	25.2	15.9	22.9	17.2	31.6	4.68	8.11
	Ours(loc)	13.4	42.3	19.8	38.6	19.4	41.1	11.7	29.4	4.13	5.84
	Ours(GC-loc)	11.5	33.6	17.7	30.2	15.1	29.7	18.9	38.6	3.27	4.10
	Ours(VL-loc)	15.1	40.1	15.0	36.2	14.9	36.5	19.8	40.2	2.40	2.19
	Ours(UM-loc)	14.4	40.2	17.9	38.4	19.6	36.8	18.3	34.5	2.22	2.42