notesum.ai

Published at November 26

LHPF: Look back the History and Plan for the Future in Autonomous Driving

cs.RO

cs.CV

Released Date: November 26, 2024

Authors: Sheng Wang¹, Yao Tian¹, Xiaodong Mei¹, Ge Sun¹, Jie Cheng¹, Fulong Ma², Pedro V. Sander¹, Junwei Liang³

Aff.: ¹HKUST; ²HKUST(GZ); ³HKUST, HKUST(GZ)

Arxiv: http://arxiv.org/abs/2411.17253v1

Refer to caption

Reactive Results
Type	Planner	Score	Collisions	Drivable	Progress	Comfort	Speed	Direction
Expert	Log-Replay	81.24	86.38	98.07	98.99	99.27	96.47	99.13
Rule-based	IDM [30]	79.31	90.92	94.04	86.16	94.40	97.33	99.63
Rule-based	PDM-Closed [14]	93.20	98.67	99.63	90.48	95.50	99.83	100.00
Pure Learning	PDM-Open [14]	54.86	81.47	87.89	69.86	99.54	97.72	97.57
	GC-PGP [17]	54.91	84.77	89.72	57.75	90.00	99.20	98.44
	RasterModel [3]	64.66	87.11	83.11	79.73	85.59	98.00	96.42
	UrbanDriver [26]	64.87	82.16	90.83	86.42	99.08	89.81	97.25
	PlanTF [8]	77.07	95.69	97.06	76.72	93.94	98.67	99.45
	PLUTO (w/o post.) [9]	78.76	97.43	97.89	75.08	96.33	98.89	99.86
	LHPF† (ours)	81.25	95.82	99.63	78.59	96.42	98.60	99.45
Non-reactive Results
Expert	Log-Replay	93.68	98.76	98.07	98.99	99.27	96.47	99.13
Rule-based	IDM [30]	79.31	90.92	94.04	86.16	94.40	97.33	99.63
Rule-based	PDM-Closed [14]	93.08	98.07	99.82	92.13	95.52	99.83	100.00
Pure Learning	PDM-Open [14]	50.24	74.54	87.89	69.86	99.54	97.72	97.57
	GC-PGP [17]	61.09	85.87	89.72	60.32	90.00	99.34	98.67
	RasterModel [3]	66.92	86.97	85.04	80.60	81.46	98.03	96.42
	UrbanDriver [26]	67.72	85.60	90.83	80.83	100.00	91.58	98.49
	PlanTF [8]	85.30	94.13	96.79	89.83	93.67	97.78	98.85
	PLUTO (w/o post.) [9]	89.04	96.18	98.53	89.56	96.41	98.13	99.49
	LHPF† (ours)	88.81	94.77	99.72	90.90	96.88	97.77	99.08