EOARD Grant

Integrating Language and Cognition in Grounded Adaptive Agents

EOARD Grant # 053060

Summary Objectives & Work plan Background Results Publications & References

Scaling up of lexicon and action repertoire in Cognitive Agents

Description:

Incremental Feature – lexicon acquisition

In the first simulation we have proposed the use of the multi-dimensional MFT in order to categorize 112 different actions. At this stage we wanted to explore the integration of language and cognition in cognitive robotic studies. Here we extend the multi-dimensional MFT algorithm, used in Simulation 1, to enable the agents to learn a lexical item to name each previous action. After performing the action, the agents will start to describe this action as three letter word. (consonant – vowel - consonant) see example table 1. Each word is unique to the action performed. This phonetic feature is dynamically added immediately after the action. At timestep 12500, (half of the training time) both features are considered when computing the fuzzy similarities. From timestep 12500, the dynamics of the σ2 fuzziness value is initialized, following equation (1), whilst σ1 continues its decrease pattern started at timestep 0. Results in figure 1 show that the model is able to categorize an action and assign a ‘word’ to this action.

Results:

Figure 1: Time evolution of the fields using as input the action and phonetic feature: 112 different actions + 112 words

Actions	Words	Actions	Words	Actions	Words	Actions	Words
1	XUS	29	WED	57	LAC	85	KUN
2	GAP	30	JIL	58	BIF	86	BAM
3	HES	31	SAK	59	FUF	87	TOQ
4	LET	32	BAR	60	HOR	88	NOK
5	PAQ	33	JOC	61	QEB	89	TIL
6	PAL	34	FUV	62	WAH	90	HAW
7	VEH	35	MEL	63	SUD	91	JOJ
8	WIG	36	JAX	64	KAW	92	WOT
9	FUH	37	JAS	65	SAV	93	REW
10	TEP	38	VUS	66	QEQ	94	FUN
11	LUT	39	KEK	67	GES	95	QIH
12	COF	40	MIQ	68	WEJ	96	KUM
13	QUT	41	PAJ	69	KEJ	97	VEF
14	TUW	42	DOJ	70	LUC	98	NUH
15	REV	43	HER	71	KAC	99	GAC
16	QER	44	SAR	72	GUQ	100	KAC
17	BIJ	45	TOD	73	PUB	101	GUG
18	QUR	46	COC	74	RIV	102	QEJ
19	QUX	47	KAH	75	DEM	103	FUG
20	XAM	48	BIG	76	BER	104	KOG
21	BOX	49	QUW	77	WIP	105	VEX
22	QIH	50	PIM	78	PUP	106	WAK
23	PUV	51	DET	79	SIK	107	WOW
24	FAC	52	FUJ	80	XIT	108	BER
25	SIL	53	MIK	81	VEN	109	CEB
26	TAX	54	MEG	82	FIP	110	TOK
27	HUT	55	RAQ	83	SIC	111	TES
28	DOK	56	MAK	84	WUW	112	LOP

Table 1: Table containing examples of words generated by the agents for a specific action

Figure 2: Teacher and learner before action is learnt and after with the addition of the ‘word’. For visualization purposes, the word is added on the image.

Up - Back