Table 1.

Reconstruction task: comparison of multi-contact prediction methods and dataset for predicting the displacements D from simulated sensors S.

Displacement Prediction Error [μm]
LR	Train	I				I and II				I, II, and III				I, II, III, and IV
		2.2				6.2				10.9				13.6
	Test	I	II	III	IV	I	II	III	IV	I	II	III	IV	I	II	III	IV
		2.2	36.6	60.3	78.2	1.9	16.1	24.4	31.8	2.0	15.6	23.0	29.8	2.1	15.6	22.6	29.0
k-NN	Train	I				I and II				I, II, and III				I, II, III, and IV
		2.3				12.8				27.5				40.9
	Test	I	II	III	IV	I	II	III	IV	I	II	III	IV	I	II	III	IV
		2.3	159.0	225.1	286.9	2.4	36.3	97.8	144.6	2.3	36.8	75.8	109.6	2.3	37.8	73.36	99.8
FNN	Train	I				I and II				I, II, and III				I, II, III, and IV
		0.6				2.9				5.7				8.1
	Test	I	II	III	IV	I	II	III	IV	I	II	III	IV	I	II	III	IV
		0.6	82.3	123.6	158.1	0.7	7.7	30.6	48.6	0.8	7.0	15.5	25.7	1.1	6.8	13.4	20.2

I, II, III, IV stand for dataset of single-, double-, triple-, and quadruple-contact, respectively. Gray columns indicate extrapolation to unseen number of contact points.