Table 20 LSTM and GRU model performance in full workspace.

From: Path-based evaluation of deep learning models for solving inverse kinematics in a revolute-prismatic robot

Model

Architecture (hidden layers × units)

Epochs

Validation loss

LSTM

3 × 64

300

0.02603

4 × 64

300

0.0159

5 × 64

300

0.1289

6 × 64

300

0.53386

GRU

3 × 64

1000

0.05792

Back to article page