Search

Dexterous in-hand manipulation for a multi-fingered anthropomorphic hand is extremely difficult because of the high-dimensional state and action spaces, rich contact patterns between the fingers and objects. Even though deep reinforcement learning has made moderate progress and demonstrated its strong potential for manipulation, it is still faced with certain challenges, such as large-scale data collection and high sample complexity. Especially, for some slight change scenes, it always needs to re-collect vast amounts of data and carry out numerous iterations of fine-tuning. Remarkably, humans can quickly transfer learned manipulation skills to different scenarios with little supervision. Inspired by human flexible transfer learning capability, we propose a novel dexterous in-hand manipulation progressive transfer learning framework (PTL) based on efficiently utilizing the collected trajectories and the source-trained dynamics model. This framework adopts progressive neural networks for dynamics model transfer learning on samples selected by a new samples selection method based on dynamics properties, rewards and scores of the trajectories. Experimental results on contact-rich anthropomorphic hand manipulation tasks show that our method can efficiently and effectively learn in-hand manipulation skills with a few online attempts and adjustment learning under the new scene. Compared to learning from scratch, our method can reduce training time costs by 95%., Comment: 12 pages, 7 figures, submitted to TNNLS

In this paper, a new robust image hashing scheme is proposed base on discrete trinion Fourier transform(DTFT). The key features of the present scheme rely on (i) The DTFT provides an effective way to jointly deal with the three channels of color images. (ii) The final hash is generated according to the amplitudes of these DTFT coefficients. (iii) Scrambling is applied at the last stage of hash generation to enhance the system security. The experiment results show that the proposed method offers good robust to most image content preserving operations and have good discrimination performance.

Object detection in thermal images is an important computer vision task and has many applications such as unmanned vehicles, robotics, surveillance and night vision. Deep learning based detectors have achieved major progress, which usually need large amount of labelled training data. However, labelled data for object detection in thermal images is scarce and expensive to collect. How to take advantage of the large number labelled visible images and adapt them into thermal image domain, is expected to solve. This paper proposes an unsupervised image-generation enhanced adaptation method for object detection in thermal images. To reduce the gap between visible domain and thermal domain, the proposed method manages to generate simulated fake thermal images that are similar to the target images, and preserves the annotation information of the visible source domain. The image generation includes a CycleGAN based image-to-image translation and an intensity inversion transformation. Generated fake thermal images are used as renewed source domain. And then the off-the-shelf Domain Adaptive Faster RCNN is utilized to reduce the gap between generated intermediate domain and the thermal target domain. Experiments demonstrate the effectiveness and superiority of the proposed method., Comment: 5 pages, 4 figures

The biological consequences of cellular senescence and immortalization in aging and cancer are in conflict. Organisms have developed common cellular signaling pathways and surveillance mechanisms to control the processing of aging against tumorigenesis. The imbalance of any signals involved in this process may result in either premature aging or tumorigenesis and reduce the life span of the organism. In contrast, the balance between aging and tumorigenesis at a higher level (homeostatic-balance) may benefit the organism with tumor-free longevity. The focus of this perspective is to review the literature on the balance between "Yin" and "Yang" in traditional Chinese medicine. Modern cellular and molecular techniques now permit a more robust system to screen herbs in traditional Chinese medicine for their activity in balancing aging and tumorigenesis. The understanding of the crosstalk between aging and tumorigenesis and new perspectives on the application of Chinese medicine might shed light on anti-aging and tumor-free strategies. J. Cell. Biochem. 118: 2581-2586, 2017. © 2017 Wiley Periodicals, Inc.

The emerging mobile crowd sensing has become a new paradigm where a crowd of mobile users utilize their smart devices to conduct complex computation and sensing tasks in mobile social networks, in which the content sharing plays a significant role. In this work, we propose a novel many-to-many content sharing framework by enabling users to exchange information through separate connections with different partners of multiple relationships simultaneously. Under the statistical channel statement information condition, the many-to-many pairing problem is formulated as a Stable Multiple Activity (SMA) matching game and solved by a distributive stable b-matching algorithm in multigraphs. Based on the matching model, we further perform the power allocation scheme to improve system performance. Simulation results demonstrate the superiority of our proposed SMA method with lower complexity.

In this study, a coupling model of agricultural and urban non-point source (AUNPS) is established in order to estimate complex non-point source (NPS) pollutant loads in the urban–rural fringe. Furthermore, a spatial–temporal change model of non-point source (CA-AUNPS) is established by the coupling of AUNPS and cellular automata (CA) models. The total nitrogen (TN) and total phosphorus (TP) loads in Tangxun watershed were simulated during 1991–2020, and the results show that: (1) the CA-AUNPS model is an effective tool to simulate the spatial–temporal changes of NPS pollutant loads in urban–rural fringe; (2) in terms of the spatial changes, the TN and TP loads generally showed as rural/urban construction land > farmland > forest/green land, and the high-value areas of NPS pollutant loads expanded from north to south with the increase of construction land; with regard to temporal changes, the TN and TP loads generally displayed an increasing tendency during 1991–2020, and by 2020 the TN and TP loads will increase to 370.06 and 33.89 t yr−1, respectively; (3) the spatial–temporal changes of TN and TP loads in Tangxun watershed correlate strongly with the land-use, slope, and rainfall.

Recently, biodegradable plastics (BPs) are emerging as a sustainable alternative to traditional plastics. When released into an aquatic environment, the biodegradable performance of BPs is influenced by biochemical processes, especially the developed plastisphere. However, studies addressing the biodegrading capacity of BPs and traditional plastics within the plastisphere are still limited. Here, we investigated plastisphere community variations and their capacity to biodegrade polyethylene terephthalate (PET) and starch-based plastics (SBP) for four time periods (15, 30, 45, and 80 days) in three freshwaters. Unexpectedly, there is no significant difference in the microbial communities and network structure of the plastisphere between SBP and PET. Moreover, SBP tended to age rapidly at the early stage (0–15 days), while the aging degree of SBP and PET did not display an obvious difference at 80 days. Partial least squares path modeling suggested that plastic aging was mainly dominated by keystone taxa of network and aquatic environmental factors. These results suggest that the aging rate of commercial BPs may not be as fast as we imagine in freshwaters (SBP ≈ PET), and the environmental behaviors of BPs in the aquatic environment should be paid more attention to. Refereed/Peer-reviewed