International Journal of Humanoid Robotics

This paper discusses about the modern techniques implemented in the field of agriculture which shaped the traditional farming to the smart farming (Agriculture 4.0). The rapid rise in the global population is one among the main reasons, besides monitoring of crops health and yield requires a huge labor force, which is also another reason for promoting intelligent systems into agriculture sector. Conventional farming methods are not suitable to meet this demand, which led robotics to associate with on-field agriculture by means of robot-based technologies like wheeled robots, ground vehicles (manned and unmanned) and aerial vehicles (manned and unmanned) that led to explore the possible advancement in agriculture 3.0. Currently, the evolutionary techniques, such as Artificial Intelligence (AI) and IoT, are implemented in robotic vehicles to make them intelligent systems. Due to unprecedented climatic change and polluted ground water for the past few decades, the crops are being infested with new varieties of diseases. This requires new image processing techniques to classify the diseases based on color, texture and, shape of leaves. The incorporation of image processing technique into AI aids in deciding the appropriate amount of herbicide supplement to the plant based on the prediction of plant growth.

To address the issue of error-prone and unstable trajectory tracking and dynamic obstacle avoidance of mobile robots in locally observable nonlinear dynamic settings, a deep reinforcement learning (RL)-based visual perception, and decision-making system is proposed. The technique creates a closed loop between the system’s environmental perception and decision-making capabilities by combining the perceptual capabilities of convolutional neural networks with the decision-making capabilities of RL in a generic form. It achieves direct output control from the visual perception input of the environment to the action through end-to-end learning. The simulation results show that this approach is capable of meeting the demands of multi-task intelligent perception and decision making. It also more effectively addresses issues with traditional algorithms, including their tendency to fall into local optimums, oscillate in groups of similar obstacles without recognizing the path, oscillate in tight spaces and inaccessible targets close to obstacles and significantly enhance real-time and adaptability of robot trajectory tracking and dynamic obstacle avoidance.

As a physiological signal reflecting the state of muscle activation, surface electromyography (sEMG) plays a vital role in the assessment of neuromuscular health, human–computer interaction, and gait analysis. Inspired by the audio signal analysis outcome that features extracted with Mel Frequency Cepstral Coefficient (MFCC) empower better representation, this paper proposes a comparative study of a gesture recognition method by using and improving with the MFCC features of sEMG. Comparing and combining with the conventional time-domain and frequency-domain features, different learning-based techniques are deployed to evaluate the performance of the proposed approach on the NinaPro datasets. The proposed approach was evaluated on the NinaPro-DB1 and NinaPro-DB2 datasets, achieving the improvements of 3.42% and 3.67%, respectively, in terms of the highest accuracy using the standard MFCC method. Correspondingly, when combined with the improved MFCC, the accuracy was further increased, reaching the maximum values of 89.82% and 87.82%, respectively, on the two datasets. The impact on the performance reveals the effectiveness of MFCC, and the results show that the proposed method has the potential to realize high-precision gesture recognition.