Ted Tao

Credit assigned CMAC and its application to online learning robust controllers

In this paper, a novel learning scheme is proposed to speed up the learning process in cerebellar model articulation controllers (CMAC). In the conventional CMAC learning scheme, the correct numbers of errors are equally distributed into all addressed hypercubes, regardless of the credibility of the hypercubes. The proposed learning approach uses the inverse of learned times of the addressed hypercubes as the credibility (confidence) of the learned values, resulting in learning speed becoming very fast. To further demonstrate online learning capability of the proposed credit assigned CMAC learning scheme, this paper also presents a learning robust controller that can actually learn online. Based on robust controllers presented in the literature, the proposed online learning robust controller uses previous control input, current output acceleration, and current desired output as the state to define the nominal effective moment of the system from the CMAC table. An initial trial mechanism for the early learning stage is also proposed. With our proposed credit-assigned CMAC, the robust learning controller can accurately trace various trajectories online.

Robust CMAC control schemes for dynamic trajectory following

Abstract Improved robust CMAC control schemes are proposed for tracing dynamic trajectories in this paper. There are two main structures in the proposed control schemes: one is the robust controller and the other is the improved CMAC network. The robust controller technique can achieve a certain goal without concern for instability of the controlled system in the presence of significant plant uncertainties if the nominal parameter is roughly estimated. Next, in order to reduce the tracing error, a suitable nominal parameter needs to be chosen. Thus, the improved CMAC learning approach under the robust control structure, using the concept of credit assignment, will be employed to determine control variables that can trace other states repeatedly during control processes. Finally, simulation results demonstrate the capability of the proposed control schemes to trace dynamic trajectories.

Closed-loop method to improve image PSNR in pyramidal CMAC networks

A closed-loop method to improve image the peak signal to noise ratio (PSNR) in pyramidal cerebellar model arithmetic computer (CMAC) networks is proposed in this paper. We propose a novel coding procedure, which can make the CMAC network learn the feature of the transmitted image with only one-shot training, so some sampled data of the original image can quickly be sent to reconstruct a coarse image. In the meantime, differential codes are transmitted to improve the image quality using the closed-loop method in pyramidal CMAC networks. As a result, the quality of the reconstructed image can be improved at the bottom of the pyramidal CMAC networks. Finally, the experimental results demonstrate that the proposed method can give higher PSNR at a lower bit rate after reconstruction, when it is applied to JPEG compression.

The treatment of image boundary effects in CMAC networks

This paper proposes a novel coding procedure which makes two dimension CMAC networks learn the feature of transmitted image without boundary effects. As a result, the reconstructed image from the proposed CMAC networks gets higher PSNR. Finally, we apply CMAC networks to JPEG compression. Experimental results demonstrate that the presented method can get higher PSNR at lower bit rate after reconstruction.

The one‐time learning hierarchical CMAC and the memory limited CA‐CMAC for image data compression

Abstract Two methods to compress transmitted image data are proposed in this paper. The first method is the one‐time learning hierarchical CMAC method and the second is the memory limited CA‐CMAC method for image data compression and reconstruction. The one‐time learning hierarchical CMAC method is used when a coarse image needs to be sent to the receiver initially and then the image quality is gradually improved at the request of the receiver. But, when the transmitting channel data is limited, the memory limited CA‐CMAC method can be used to decrease the bit rate per pixel. Both proposed methods, unlike conventional compression methods, use no filtering technique in either compression or reconstruction. CMAC networks use a few hypercubes to learn the characteristics of the original image, so image data can be compressed without suffering from blocking effect or boundary effect. Onetime learning is good enough for compressing image data, and it has a high SNR after reconstruction.

The auxiliary CMAC applied to online tuning robust fuzzy controllers

A novel auxiliary CMAC robust fuzzy control schemes is proposed in this paper. There are two structures in the proposed schemes: one is the robust fuzzy controller and the other is the auxiliary CMAC learning algorithm. The robust fuzzy controller can achieve a certain goal without concern for instability of the controlled system in the presence of significant plant uncertainties, if the nominal parameter is roughly estimated. In order to improve the performance of robust fuzzy controller, the nominal parameter should be adjusted. Thus, an auxiliary CMAC learning algorithm under the robust fuzzy control structure is employed to online tuning the nominal parameter. Finally, simulation results demonstrate the excellent capability of the proposed structure for improving the output performance.

Adaptive Fuzzy Sliding Mode Control Schemes for Tracking Time-various Trajectories

Adaptive fuzzy sliding mode control schemes are proposed for tracking time-various trajectories in this paper. There are two main structures in the proposed control schemes: one is the sliding mode controller and the other is the adaptive fuzzy schemes. The sliding mode controller ensures the Lyapunov stability of the system, if the reference signal and all signals involved are bound. Next, the adaptive fuzzy schemes can improve chattering and delay phenomena in the sliding mode control system. Finally, simulation results confirm that the proposed adaptive fuzzy sliding mode control schemes can track time-various trajectories quickly and accurately.

Differential Codes Transmitting in Pyramidal CMAC Networks

Differential codes transmitting in pyramidal CMAC networks to improve image PSNR is proposed in this paper. We propose a novel coding procedure, which can make the CMAC network use few computations to learn the feature of transmitted image with only one-shot training. So some sampled data of the original image can be sent to reconstruct a coarse image quickly. In the meanwhile, differential codes are transmitted in the proposed pyramidal architecture by the request of receiver to improve the image quality. As a result, the reconstructed image at the bottom of pyramidal CMAC networks will get higher PSNR. Finally, pyramidal CMAC networks are applied to JPEG compression. Experimental results demonstrate the proposed method can get higher PSNR at lower bit rate after reconstruction.

The CA-CMAC for downsampling image data size in the compressive domain

The CA-CMAC for downsampling image data size in the compressive domain is proposed in this paper. When the transmitting data is limited, it can reduce the bit rate during transmitting image data and decrease computations per pixel during the reconstructive process. The proposed method maps the image data into the CMAC lookup table, which can learn the characteristics of original image and can change image data size during downsampling and upsampling processes. It is unlike the conventional linear interpolation method, which gets lower SNR and costs more computation in the compression and reconstructive processes. The CA-CMAC method uses only a few hypercubes to learn the characteristics of original image, and transmits the learned characteristics to the receiver for reconstruction. Finally, the proposed method is applied to downsample JPEG data size in this paper, and it is shown that it gets high SNR after reconstruction.

The CMAC based FLC and its application to rear-loading truck problems

In this paper, the CMAC-based FLC is applied to rear-loading truck problems. We show that the proposed CMAC-based FLC has fast learning capability in forward and backward computations because it possesses the local generalization and because it has only a small number of activated units (hypercubes) in the network. Simulation results validate the fast learning and the accurate approximation of the proposed CMAC-based FLC.