Research on UUV Obstacle Avoiding Method Based on Recurrent Neural Networks

Table 1

The performance of all structures in training.

Structure	CW-RNN96	CW-RNN180	LSTM18	LSTM45
Number of units in hidden layer	Rb=24	Rb=45	Lb=18	Lb=45
Number of parameters in hidden layer	6177(mean)	17253(mean)	5832	19440
Mean train time of each iteration/s	6.36	11.85	6.44	18.29
Initial error	0.454	0.645	2.800	4.600
Best MSE	1.74×10^−3	1.26×10^−3	1.52×10^−3	8.7×10^−4
Convergence generation	15000	18000	9500	13500