Proceedings of National Conference on IT- Contemporary and Future Technologies for Social Change ITM universe, Gwalior.11-12 March 2011. Paper ID: 3402 1 Abstract— This paper presents the comprehensive overview of the recent developments in hybrid artificial intelligence and control techniques, their structure, properties and their respective fields of implementation. This paper surveys all the neural techniques developed in last decade together with boundaries of various techniques and identifies the importance of using these techniques. Index Terms—control techniques, hybrid neural, new trend in AI I. INTRODUCTION There are various Artificial intelligence techniques which are used now days in various fields of control, whether in industries, plants, and robotics .But in the competition of developing more efficient techniques for control the combinations of various techniques are going on to be used. Three AI techniques were widely applied: Artificial Neural Networks (ANNs) [23], Fuzzy Logic Control (FLC) [24] and Genetic Algorithms (GA) [25]. Merging more than one AI technique is common now days. These techniques are quite efficient in their field of use. This paper may prove to lead the designers to ease in making choice of techniques for particular type of control scheme. This paper surveys all the AI techniques developed by last decade and their fields of implementation. The techniques like ANFIS[1], GFIS[36], CANFIS, SANFIS[33,34], MANFIS[35], ANN, MEFFN, PRNN, GARIC[25], FALCON[26], NEFCON[27], FUN[28], SONFIN[29], FINEST[30], EFuNN[31], dmEFuNN[31], evolutionary design of neuro fuzzy systems[32], AEN, ASN, NEFPROX, FNN[2], SOFNN [3], SOFNNGA[19], RAN[4], RBFN,AR-RBFN[5],RB-FAFS[37], DFNN[6], GDFNN[7] , Farag’s model[20], RBFNF, AFNC[38], NiF-T[21], RNN[21], ORNN[21], GBFLC[17], ASAFES2, R-SANFIS[10], MSSC[13], SLA[15], MSF, GAMSF, FUZZY NAV,NEURO-NAV,ALVINN,ALVINN-VC, PCs, GCs, SGCs, DGCs, FGCs, GRBFs, ACs, PVCs, BVCs,WCS, ACS, ECS, SLN, NEFCLASS, NFS, CTRNN, FSOM, NFLC[22], GRBFs,FGNN,MSFC[8],MV-ANFC[9] has variety of Manuscript received January 27, 2011. Umesh Kumar soni is persuing M.E. in Control System in Jabalpur Engineering college Jabalpur.( phone: 9407302484; e-mail: soni.umesh9@ gmail.com). Prof. Hemant Amhia is with Electrical Engineering Department,Jabalpur Engineering College, Jabalpur, M.P., India ( phone: 9827302392, e- mail:hemant_dreamzin@yahoo.co.in) matching structure and properties.Out of all the above techniques latest techniques are discussed in detail as follows. [1] SOFNNGA: self organizing fuzzy neural network based on genetic algorithm is based on a genetic algorithm to design a fuzzy neural network, named self-organizing fuzzy neural network based on GAs (SOFNNGA) [19], to implement Takagi–Sugeno (TS) fuzzy models. One of the main novelties of the proposed approach is that the model is built for a system without a priori knowledge about the partitions of input space and the number of fuzzy rules. The SOFNNGA is based on ellipsoidal basis function (EBF) neurons consisting of a center vector and a width vector. Each EBF neuron represents a partition in the input space. The premise part (i.e., IF part) of a fuzzy rule is derived from the EBF neuron. The weighted bias in the weighted layer is a consequence part (i.e., THEN part) of a fuzzy rule. There are two stages in this algorithm. In the first stage, the initial structure is generated from an empty set of EBF neurons without a priori knowledge of how to partition the input space. The GA is introduced in the second stage and attempts to identify the least important EBF neurons and deletes those neurons to yield a compact structure. The architecture of the SOFNNGA is a five layer fuzzy neural network shown in Fig.1. The structure is the same as SOFNN in [3]. Fig. 2 illustrates the internal structure of the jth neuron in the EBF layer. Layer 1 is the input layer. Each neuron in this layer represents an input variable. Layer 2 is the EBF layer. Each neuron in this layer represents an IF part (or premise) of a fuzzy rule. Layer 3 is the normalized layer. The number of neurons in this layer is equal to that of layer 2. Layer 4 is the weighted layer. Each neuron in this layer has two inputs which are the output of the related neuron in the layer 3 and the weighted bias w2j. Layer 5 is the output layer. Fig.1 Structure of SOFNNGA. The Overview of New Trends in Hybrid Artificial - Intelligence and Control Techniques Umesh Kumar Soni, Prof. Hemant Amhia