In the Industry 4.0, edge industrial services such as smart robotic services are widely used in smart factory. The workflow of these services mainly consists of task decomposition and resource allocation. The long scheduling time, high communication delay and load imbalance among edge nodes are the challenging problems. Traditional cloud manufacturing platforms are difficult to meet the new requirements. It is hard for the existing scheduling methods to maintain a balance between algorithm complexity and performance. Training scheduling data by deep learning has become a feasible method to achieve fast prediction of the scheduling results. In this paper, a cloud edge-based two-level hybrid scheduling learning model is put forward at first. Then an improved bat scheduling algorithm with interference factors and variable step size (VSSBA) is proposed. And then, according to the historical scheduling data, the improved long and short-term memory networks (LSTM) model is put forward for fast prediction of the cloud-edge collaborative scheduling results. Experiments show that our proposed learning model can improve the performance of the cloud manufacturing platform in real-life applications efficiently. Finally, future research issues and challenges are identified.

Edge computing has strong real-time and big data interaction processing requirements. The long scheduling time and load imbalance among edge nodes and edge servers are the key problems of edge computing. The current cloud computing scheduling algorithms all have balance problems between algorithm complexity and performance, and cannot fundamentally solve the contradiction. It is a feasible method to use the deep learning model to train the scheduled data to achieve direct prediction of the scheduling results. This paper mainly studies from two aspects, one is to obtain more accurate training data from the perspective of researching optimal scheduling algorithms, and the other is to improve the training speed from the perspective of improving the deep learning model. At first an improved chaotic bat swarm algorithm is put forward. It introduces chaotic factors and second-order oscillation mechanisms to improve the speed of update and dynamic parameter mechanisms. And then the long short-term memory network deep learning model is trained with the historical data by the improved algorithm. The experimental results show that the improved learning model can achieve the purpose of quickly predicting the scheduling result.

Mobile terminal gesture recognition is an extreme challenge, not only because its limited computing resource make it complicated to identify feature points but also the complex background can easily affect the recognition result. This study proposes a gesture recognition method based on improved features from accelerated segment test (FAST) corner detection. First, in order to eliminate the effects of complex background and light, the intersection of the two frame images is obtained through background subtraction and the multi‐colour space to realise the detection of the hand. Second, in order to improve the performance of the algorithm, an improved FAST corner detection method combined with the back propagation neural network (BPNN) is proposed in accordance with the characteristics of fingertips. Subsequently, the feature points are screened by method of non‐maximum suppression. Finally, gesture recognition is realised by matching feature points. Experimental results illustrate that this method has strong anti‐interference ability in complex background, and it is good at performance.