A Fusion Method of Local Path Planning for Mobile Robots Based on LSTM Neural Network and Reinforcement Learning

Table 6

Experiments with different numbers of hidden layers and neuron nodes.

No.	Hidden layer	Training set			Test set			Success rate
		R^2	RMSE	Loss	R^2	RMSE	Loss
1	L128	0.9558	0.0324	0.0011	0.8749	0.0571	0.0033	86%
2	L150	0.9537	0.0332	1.10E − 03	0.8772	0.0566	0.0032	91%
3	L100 F50	0.9417	0.0372	1.40E − 03	0.8727	0.0576	0.0033	88%
4	L128 F64	0.9468	0.0356	0.0013	0.8796	0.056	0.0031	97%
5	L100 L50	0.978	0.0229	5.23E − 04	0.8721	0.0577	0.0033	89%