Low-Cost Recognition of Plastic Waste Using Deep Learning and a Multi-Spectral Near-Infrared Sensor.

This work presents an approach for the recognition of plastics using a low-cost spectroscopy sensor module together with a set of machine learning methods. The sensor is a multi-spectral module capable of measuring 18 wavelengths from the visible to the near-infrared. Data processing and analysis are performed using a set of ten machine learning methods (Random Forest, Support Vector Machines, Multi-Layer Perceptron, Convolutional Neural Networks, Decision Trees, Logistic Regression, Naive Bayes, k-Nearest Neighbour, AdaBoost, Linear Discriminant Analysis). An experimental setup is designed for systematic data collection from six plastic types including PET, HDPE, PVC, LDPE, PP and PS household waste. The set of computational methods is implemented in a generalised pipeline for the validation of the proposed approach for the recognition of plastics. The results show that Convolutional Neural Networks and Multi-Layer Perceptron can recognise plastics with a mean accuracy of 72.50% and 70.25%, respectively, with the largest accuracy of 83.5% for PS plastic and the smallest accuracy of 66% for PET plastic. The results demonstrate that this low-cost near-infrared sensor with machine learning methods can recognise plastics effectively, making it an affordable and portable approach that contributes to the development of sustainable systems with potential for applications in other fields such as agriculture, e-waste recycling, healthcare and manufacturing.

A Light-Weight Artificial Neural Network for Recognition of Activities of Daily Living.

Human activity recognition (HAR) is essential for the development of robots to assist humans in daily activities. HAR is required to be accurate, fast and suitable for low-cost wearable devices to ensure portable and safe assistance. Current computational methods can achieve accurate recognition results but tend to be computationally expensive, making them unsuitable for the development of wearable robots in terms of speed and processing power. This paper proposes a light-weight architecture for recognition of activities using five inertial measurement units and four goniometers attached to the lower limb. First, a systematic extraction of time-domain features from wearable sensor data is performed. Second, a small high-speed artificial neural network and line search method for cost function optimization are used for activity recognition. The proposed method is systematically validated using a large dataset composed of wearable sensor data from seven activities (sitting, standing, walking, stair ascent/descent, ramp ascent/descent) associated with eight healthy subjects. The accuracy and speed results are compared against methods commonly used for activity recognition including deep neural networks, convolutional neural networks, long short-term memory and convolutional-long short-term memory hybrid networks. The experiments demonstrate that the light-weight architecture can achieve a high recognition accuracy of 98.60%, 93.10% and 84.77% for seen data from seen subjects, unseen data from seen subjects and unseen data from unseen subjects, respectively, and an inference time of 85 µs. The results show that the proposed approach can perform accurate and fast activity recognition with a reduced computational complexity suitable for the development of portable assistive devices.

Towards Living Machines: current and future trends of tactile sensing, grasping, and social robotics.

The development of future technologies can be highly influenced by our deeper understanding of the principles that underlie living organisms. The Living Machines conference aims at presenting (among others) the interdisciplinary work of behaving systems based on such principles. Celebrating the 10 years of the conference, we present the progress and future challenges of some of the key themes presented in the robotics workshop of the Living Machines conference. More specifically, in this perspective paper, we focus on the advances in the field of biomimetics and robotics for the creation of artificial systems that can robustly interact with their environment, ranging from tactile sensing, grasping, and manipulation to the creation of psychologically plausible agents.

Wearable Assistive Robotics: A Perspective on Current Challenges and Future Trends.

Wearable assistive robotics is an emerging technology with the potential to assist humans with sensorimotor impairments to perform daily activities. This assistance enables individuals to be physically and socially active, perform activities independently, and recover quality of life. These benefits to society have motivated the study of several robotic approaches, developing systems ranging from rigid to soft robots with single and multimodal sensing, heuristics and machine learning methods, and from manual to autonomous control for assistance of the upper and lower limbs. This type of wearable robotic technology, being in direct contact and interaction with the body, needs to comply with a variety of requirements to make the system and assistance efficient, safe and usable on a daily basis by the individual. This paper presents a brief review of the progress achieved in recent years, the current challenges and trends for the design and deployment of wearable assistive robotics including the clinical and user need, material and sensing technology, machine learning methods for perception and control, adaptability and acceptability, datasets and standards, and translation from lab to the real world.

A model identification approach to quantify impact of whole-body vertical vibrations on limb compliant dynamics and walking stability.

Extensive research is ongoing in the field of orthoses/exoskeleton design for efficient lower limbs assistance. However, despite wearable devices reported to improve lower limb mobility, their structural impacts on whole-body vertical dynamics have not been investigated. This study introduced a model identification approach and frequency domain analysis to quantify the impacts of orthosis-generated vibrations on limb stability and contractile dynamics. Experiments were recorded in the motion capture lab using 11 unimpaired subjects by wearing an adjustable ankle-foot orthosis (AFO). The lower limb musculoskeletal structure was identified as spring-mass (SM) and spring-mass-damper (SMD) based compliant models using the whole-body centre-of-mass acceleration data. Furthermore, Nyquist and Bode methods were implemented to quantify stabilities resulting from vertical impacts. Our results illustrated a significant decrease (p < 0.05) in lower limb contractile properties by wearing AFO compared with a normal walk. Also, stability margins quantified by wearing AFO illustrated a significant variance in terms of gain-margins (p < 0.05) for both loading and unloading phases whereas phase-margins decreased (p < 0.05) only for the respective unloading phases. The methods introduced here provide evidence that wearable orthoses significantly affect lower limb vertical dynamics and should be considered when evaluating orthosis/prosthesis/exoskeleton effectiveness.

Evaluation of gait transitional phases using neuromechanical outputs and somatosensory inputs in an overground walk.

In a bipedal walk, the human body experiences continuous changes in stability especially during weight loading and unloading transitions which are reported crucial to avoid fall. Prior stability assessment methods are unclear to quantify stabilities during these gait transitions due to methodological and/or measurement limitations. This study introduces Nyquist and Bode methods to quantify stability gait transitional stabilities using the neuromechanical output (CoP) and somatosensory input (GRF) responses. These methods are implemented for five different walking conditions grouped into walking speed and imitated rotational impairments. The trials were recorded with eleven healthy subjects using motion cameras and force platforms. The time rate of change in O/Is illustrated impulsive responses and modelled in the frequency domain. Nyquist and Bode stability methods are applied to quantify stability margins. Stability margins from outputs illustrated loading phases as stable and unloading phases as unstable in all walking conditions. There was a strong intralimb compensatory interaction (p < .001, Spearman correlation) found between opposite limbs. Overall, both walking groups illustrated a decrease (p < .05, Wilcoxon signed-rank test) in stability margins compared with normal/preferred speed walk. Further, stabilities quantified from outputs were found greater in magnitudes than the instability quantified from inputs illustrating the neuromotor balance control ability. These stability outcomes were also compared by applying extrapolated-CoM method. These methods of investigating gait dynamic stability are considered as having important implications for the assessment of ankle-foot impairments, rehabilitation effectiveness, and wearable orthoses.

Memory and mental time travel in humans and social robots.

From neuroscience, brain imaging and the psychology of memory, we are beginning to assemble an integrated theory of the brain subsystems and pathways that allow the compression, storage and reconstruction of memories for past events and their use in contextualizing the present and reasoning about the future-mental time travel (MTT). Using computational models, embedded in humanoid robots, we are seeking to test the sufficiency of this theoretical account and to evaluate the usefulness of brain-inspired memory systems for social robots. In this contribution, we describe the use of machine learning techniques-Gaussian process latent variable models-to build a multimodal memory system for the iCub humanoid robot and summarize results of the deployment of this system for human-robot interaction. We also outline the further steps required to create a more complete robotic implementation of human-like autobiographical memory and MTT. We propose that generative memory models, such as those that form the core of our robot memory system, can provide a solution to the symbol grounding problem in embodied artificial intelligence. This article is part of the theme issue 'From social brains to social robots: applying neurocognitive insights to human-robot interaction'.

Adaptive Bayesian inference system for recognition of walking activities and prediction of gait events using wearable sensors.

In this paper, a novel approach for recognition of walking activities and gait events with wearable sensors is presented. This approach, called adaptive Bayesian inference system (BasIS), uses a probabilistic formulation with a sequential analysis method, for recognition of walking activities performed by participants. Recognition of gait events, needed to identify the state of the human body during the walking activity, is also provided by the proposed method. In addition, the BasIS system includes an adaptive action-perception method for the prediction of gait events. The adaptive approach uses the knowledge gained from decisions made over time by the inference system. The action-perception method allows the BasIS system to autonomously adapt its performance, based on the evaluation of its own predictions and decisions made over time. The proposed approach is implemented in a layered architecture and validated with the recognition of three walking activities:level-ground, ramp ascent and ramp descent. The validation process employs real data from three inertial measurements units attached to the thigh, shanks and foot of participants while performing walking activities. The experiments show that mean decision times of 240 ms and 40 ms are needed to achieve mean accuracies of 99.87% and 99.82% for recognition of walking activities and gait events, respectively. The validation experiments also show that the performance, in accuracy and speed, is not significantly affected when noise is added to sensor measurements. These results show that the proposed adaptive recognition system is accurate, fast and robust to sensor noise, but also capable to adapt its own performance over time. Overall, the adaptive BasIS system demonstrates to be a robust and suitable computational approach for the intelligent recognition of activities of daily living using wearable sensors.

Prediction of gait events in walking activities with a Bayesian perception system.

In this paper, a robust probabilistic formulation for prediction of gait events from human walking activities using wearable sensors is presented. This approach combines the output from a Bayesian perception system with observations from actions and decisions made over time. The perception system makes decisions about the current gait events, while observations from decisions and actions allow to predict the most probable gait event during walking activities. Furthermore, our proposed method is capable to evaluate the accuracy of its predictions, which permits to obtain a better performance and trade-off between accuracy and speed. In our work, we use data from wearable inertial measurement sensors attached to the thigh, shank and foot of human participants. The proposed perception system is validated with multiple experiments for recognition and prediction of gait events using angular velocity data from three walking activities; level-ground, ramp ascent and ramp descent. The results show that our method is fast, accurate and capable to evaluate and adapt its own performance. Overall, our Bayesian perception system demonstrates to be a suitable high-level method for the development of reliable and intelligent assistive and rehabilitation robots.