Learning from Demonstrations and Human Evaluative Feedbacks: Handling Sparsity and Imperfection Using Inverse Reinforcement Learning Approach

Table 3

Learning the driving styles from a human teacher in different conditions.

Driving style	Demonstrations			Number of feedbacks		EV of	EV of teacher policy
	Type	Time	Optimality degree	All	Negative
Style 1	Abundant	120 sec	70%	164	62	16.821	17.164
	Sparse	20 sec	100%	302	84	16.906
	No demo	—	—	409	239	16.684
Style 2	Abundant	120 sec	70%	124	48	14.893	15.275
	Sparse	20 sec	100%	218	59	14.969
	No demo	—	—	306	182	14.877