Tensor training (TT) decomposition and Tucker decomposition are two effective compression methods for convolutional neural networks. However, TT and Tucker decomposition face the problems of spatial structure information loss and high computational complexity respectively. To solve the above problems, this paper considered the information retention rate and resource occupancy of the network structure and proposed a LC convolutional neural network compressed method (TT-LC) without pre-training based on TT-Tucker decomposition, adopting the learning-compression (LC) algorithm constraint compression framework. The TT-LC method includes two parts: learning step and compression step. The learning step didn’t not need the pre-training process, and adopted the exponential cyclic learning rate method to improve the training accuracy. In the compression step, this paper selected the global optimal rank according to the advantages of TT and Tucker decomposition and the characteristics of Bayes rule, and used empirically variable Bayesian matrix factorization (EVBMF) and Bayesian optimization (BayesOpt) to select reasonable ranks to guide tensor decomposition. The TT-LC method was used to compress the trained model. TT-LC method not only reduces the loss rate of spatial structure information and computational complexity, but also solves the problem that the unreasonable rank selection of the tensor leads to the significant decrease in model accuracy. It can realize the double Bayesian rank selection and double compression of the model, and obtains the optimal compression model. Finally, experiments were carried out on CIFAR10 and CIFAR100 datasets using ResNets and VGG networks. The results show that for ResNet32 network, compared with the benchmark method, the proposed method achieved a compression rate of parameter quantity of 69.6% and a floating point computation compression rate of 66.7% with the accuracy of 92.22%.

In order to improve the accuracy and compliance of human-robot physical interaction and achieve optimal interaction performance, a human-robot physical interaction control method based on iterative optimal impedance was proposed to solve the problem that iterative learning-based impedance control method needs to repeat the same task many times. The proposed method draws on the mechanism by which iterative optimal control can optimize cost function to determine optimal control input to the system without information of the system matrix. A double-loop control structure was used for the proposed control method. An iterative optimal impedance controller (IOIC) was designed for a task-oriented outer loop. The problem of determining optimal impedance parameters was described as a linear quadratic regulator problem, which utilized iterative optimal control to find optimal feedback gain and minimize cost function including tracking error and interaction force. Robot jitter caused by parameter mutations was avoided by introducing soft auxiliary functions. A nonsingular terminal sliding mode trajectory tracking controller (NTSMTC) was used in the inner loop of the robot to make the actual trajectory of the robot track impedance trajectory output by the outer loop, and the chattering of control law was eliminated by saturation function. Simulation results prove that the proposed method can obtain optimal impedance parameters only by using interactive information in the initial stage of the task once in a human-robot collaborative task, so as to minimize the trajectory tracking error and the force consumed by the human during the task.