HI-MAPF Results — Bharath Muppasani

HI-MAPF MAPF-GPT-2M DCC EPH SCRIMP

random-32-32-20 map

random-32-32-20

32×32 grid · 20% obstacle density · 10–128 agents

	Learning-based									Hybrid
	SCRIMP			DCC			MAPF-GPT-2M			EPH			HI-MAPF
	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓
10	100	48.0	72.8×	100	49.1	7.3×	100	49.2	142×	100	47.6	5.5×	100	50.5	3.6 (1×)
20	100	50.5	30×	100	58.0	11×	100	53.8	137×	95	62.5	11×	100	56.0	18.8 (1×)
30	100	52.4	13×	100	59.5	10.5×	100	53.9	82×	100	62.0	8.7×	100	55.5	66.3 (1×)
40	95	58.0	28.2×	95	72.8	28.8×	100	57.2	205×	100	64.5	28.7×	100	107.0	48.8 (1×)
50	100	56.8	7.4×	90	81.7	15×	100	58.9	76×	95	79.0	14×	100	61.4	221.4 (1×)
60	90	71.4	13×	70	92.0	30×	100	63.0	120×	95	88.7	29×	100	128.2	214.2 (1×)
70	90	70.5	1.5×	60	104.5	4.8×	100	68.2	17×	65	110.7	5.4×	100	65.5	2,158.2 (1×)
80	80	83.8	3.3×	60	111.7	12×	100	72.9	37×	70	108.9	14×	100	71.8	1,405.5 (1×)
90	60	96.9	3.2×	30	117.8	15×	100	70.0	34×	25	123.8	18×	95	82.8	1,950.7 (1×)
100	60	97.2	1.9×	40	117.5	10×	100	77.0	23×	50	121.0	11×	85	76.5	3,693.9 (1×)
128	70	96.6	–	0	–	–	90	94.1	–	5	127.5	–	0	–	–

random-64-64-20 map

random-64-64-20

64×64 grid · 20% obstacle density · 10–128 agents

	Learning-based									Hybrid
	SCRIMP			DCC			MAPF-GPT-2M			EPH			HI-MAPF
	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓
10	50	112.5	680×	100	97.7	12×	100	99.6	350×	95	97.3	8.7×	100	108.0	0.9 (1×)
20	10	126.5	223×	100	107.8	9.6×	100	110.2	231×	100	105.7	7.7×	100	115.6	6.2 (1×)
30	10	126.7	94×	90	112.2	7.4×	100	111.7	147×	100	109.5	6.1×	100	118.2	23.4 (1×)
40	0	–	–	100	112.8	8.5×	100	112.7	137×	100	112.3	7.3×	100	122.6	45.0 (1×)
50	0	–	–	85	116.8	8.1×	100	115.2	90×	100	113.8	5.8×	100	124.0	109.8 (1×)
60	0	–	–	65	120.8	11×	85	117.7	130×	95	116.1	8.7×	100	130.2	110.8 (1×)
70	0	–	–	50	123.5	10×	95	118.7	112×	85	120.8	9.0×	100	128.2	183.0 (1×)
80	0	–	–	20	126.8	14×	95	121.0	125×	90	120.2	12×	100	128.7	211.9 (1×)
90	0	–	–	60	124.6	17×	75	121.3	140×	80	122.2	13×	95	130.4	243.6 (1×)
100	0	–	–	25	127.2	13×	75	123.8	89×	45	125.3	11×	90	130.1	472.9 (1×)
128	0	–	–	0	–	–	55	126.2	132×	15	127.7	23×	5	132.0	530.0 (1×)

den312d map

den312d

65×81 game map · structured layout · 8–128 agents

	Learning-based									Hybrid
	SCRIMP			DCC			MAPF-GPT-2M			EPH			HI-MAPF
	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓
8	20	226.2	321×	100	114.2	6.3×	100	114.0	83×	100	111.9	3.6×	100	115.3	3.6 (1×)
16	0	–	–	100	130.7	8.8×	100	120.9	67×	100	122.2	5.9×	100	124.0	21.8 (1×)
32	0	–	–	85	163.5	11×	100	126.4	61×	100	127.0	6.0×	100	136.7	105.2 (1×)
64	0	–	–	90	190.7	11×	100	137.8	36×	100	146.1	8.8×	100	153.1	759.9 (1×)
128	0	–	–	5	254.8	–	35	234.0	–	90	207.7	–	0	–	–

warehouse map

warehouse

161×63 structured warehouse · 8–128 agents

	Learning-based									Hybrid
	SCRIMP			DCC			MAPF-GPT-2M			EPH			HI-MAPF
	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓	SR↑	MS↓	IU↓
8	0	–	–	100	191.8	5.4×	100	193.6	134×	100	191.6	2.1×	100	192.8	0.8 (1×)
16	0	–	–	100	213.3	16×	100	214.0	510×	100	212.9	11×	100	215.1	1.0 (1×)
32	0	–	–	100	219.2	8.3×	100	221.8	215×	100	218.1	5.4×	100	220.3	10.7 (1×)
64	0	–	–	100	227.6	15×	100	229.3	401×	100	227.8	11×	100	227.6	23.6 (1×)
128	0	–	–	95	236.8	14.5×	95	237.2	277×	90	239.1	10.8×	100	234.7	139.1 (1×)

Hardware Validation

5 TurtleBot4 robots in a 6×6 physical grid, ROS 2, distributed deployment. Results averaged across 5 problem instances.

Metric	SCRIMP	DCC	MAPF-GPT	EPH	HI-MAPF
Success Rate (%)	100	100	100	100	100
Avg. Makespan	16.4	19.4	13.6	21.8	12.2
Avg. Total IU	42,413	596	588	623	199
IU vs HI-MAPF	213×	3.0×	3.0×	3.1×	1×
Avg. Comm. Freq.	180	116	82	131	10
Avg. Exec Time (s)	211.2	235.0	176.2	234.9	181.5
Avg. Exec Msgs	169	199	141	223	127

Per-Instance Hardware Results

Individual results for each of the 5 problem instances on a 6×6 grid with 5 TurtleBot4 robots.

	Learning-based									Hybrid
	SCRIMP			DCC			MAPF-GPT-2M			EPH			HI-MAPF
	MS	IU	Exec(s)	MS	IU	Exec(s)	MS	IU	Exec(s)	MS	IU	Exec(s)	MS	IU	Exec(s)
Prob 1	14	37,128	185.4	18	540	214.8	12	504	162.3	20	560	218.7	11	180	168.2
Prob 2	18	45,720	225.6	20	620	241.5	14	612	182.4	22	648	240.1	13	210	190.5
Prob 3	15	39,840	198.2	19	580	228.6	13	556	170.8	21	602	228.4	12	195	175.3
Prob 4	17	44,256	218.9	21	640	248.2	15	624	185.6	23	670	245.8	13	208	188.7
Prob 5	18	45,121	228.1	19	600	241.8	14	644	180.0	23	635	241.5	12	202	184.6