R. de Azambuja, V.J. Brusamarello, S. Haffner, and R. Wolff Porto. “Analysis and Optimization of an Inductive Power Transfer With a Randomized Method.” IEEE Transactions on Instrumentation and Measurement 63, no. 5 (May 2014): 1145–52. doi:10.1109/TIM.2013.2296397.
Abstract: This paper introduces the analysis of the efficiency and transferred power of an inductive link circuit with different network configurations of capacitors connected to primary and secondary coils. The best performance for both cited objective functions was observed with two capacitors connected to the input coil and two capacitors connected to the output coil. However, the output equations in this circuit configuration for both efficiency and output power are very complex and a numerical method had to be applied to compute the capacitors values. Since an exhaustive search would be long, some simplifications were assumed to reduce the search space and the processing time. Thus, a search algorithm based on a randomized method was developed and successfully applied. The results for both efficiency and output power of four capacitors configuration were compared with other usual approaches, such as the single and two capacitors compensation. Finally, a basic prototype was built and the theoretical results were validated. Both simulated and experimental results of the four capacitor configuration showed a significant improvement on the efficiency and output power of the inductive link.
V. J. Brusamarello, Y. B. Blauth, R. de Azambuja, I. Muller, and F. R. de Sousa. “Power Transfer With an Inductive Link and Wireless Tuning.” IEEE Transactions on Instrumentation and Measurement 62, no. 5 (May 2013): 924–31. doi:10.1109/TIM.2013.2245041.
Abstract: This paper presents the analysis of two air-coupled coils used to transfer energy to charge a battery. This battery is used to power an electronic device designed to monitor variables such as impact strength, range of temperature, and humidity associated with the transport of fruits. The device is inside a sealed enclosure that cannot be opened for recharging the battery. The study shows that the coupled coils need to work with a resonance capacitor, at least on the secondary coil. However, the resonance frequency also depends on the coupling factor k. Therefore, this work proposes a monitoring system with a closed loop for fine-tuning the resonance frequency of the secondary coil circuit. Before starting charging the battery, the system scans the resonance frequency on the primary coil and measures the output power on the secondary coil looking for the optimal point. This procedure reduces problems of coupling factor variations with positioning of the coils during the battery charging.
R. de Azambuja, F. B. Klein, M. F. Stoelen, S. V. Adams, and A. Cangelosi. “Graceful Degradation Under Noise on Brain Inspired Robot Controllers.” In Neural Information Processing, edited by Akira Hirose, Seiichi Ozawa, Kenji Doya, Kazushi Ikeda, Minho Lee, and Derong Liu, 195–204. Lecture Notes in Computer Science 9947. Springer International Publishing, 2016. doi:10.1007/978-3-319-46687-3_21.
Abstract: How can we build robot controllers that are able to work under harsh conditions, but without experiencing catastrophic failures? As seen on the recent Fukushima’s nuclear disaster, standard robots break down when exposed to high radiation environments. Here we present the results from two arrangements of Spiking Neural Networks, based on the Liquid State Machine (LSM) framework, that were able to gracefully degrade under the effects of a noisy current injected directly into each simulated neuron. These noisy currents could be seen, in a simplified way, as the consequences of exposition to non-destructive radiation. The results show that not only can the systems withstand noise, but one of the configurations, the Modular Parallel LSM, actually improved its results, in a certain range, when the noise levels were increased. Also, the robot controllers implemented in this work are suitable to run on a modern, power efficient neuromorphic hardware such as SpiNNaker.
R. de Azambuja, A. Cangelosi, and S.V. Adams. “Diverse, Noisy and Parallel: A New Spiking Neural Network Approach for Humanoid Robot Control.” In 2016 International Joint Conference on Neural Networks (IJCNN), 1134–42. Vancouver, 2016. doi:10.1109/IJCNN.2016.7727325.
Abstract: How exactly our brain works is still an open question, but one thing seems to be clear: biological neural systems are computationally powerful, robust and noisy. Using the Reservoir Computing paradigm based on Spiking Neural Networks, also known as Liquid State Machines, we present results from a novel approach where diverse and noisy parallel reservoirs, totalling 3,000 modelled neurons, work together receiving the same averaged feedback. Inspired by the ideas of action learning and embodiment we use the safe and flexible industrial robot BAXTER in our experiments. The robot was taught to draw three different 2D shapes on top of a desk using a total of four joints. Together with the parallel approach, the same basic system was implemented in a serial way to compare it with our new method. The results show our parallel approach enables BAXTER to produce the trajectories to draw the learned shapes more accurately than the traditional serial one.
R. W. Porto, V. J. Brusamarello, I. Muller, F. R. Sousa, and R. Azambuja. “Design and Optimization of a Power Inductive Link.” In Instrumentation and Measurement Technology Conference (I2MTC), 2014 IEEE International, 648–53. IEEE, 2014. doi:10.1109/I2MTC.2014.6860823.
Abstract: The wireless power transfer efficiency plays an important role in design of implantable biomedical devices. There are some important constraints in the design, such as the size of the implantable coil as well as the power delivered to the implant. In this context, this paper presents the design of two planar spiral coils coupled by air used to transferring energy wirelessly. The geometric constraints of the application are determined to define the dimensions of the coils. These parameters along with the working frequency of the system are applied to a procedure that results in the equivalent electrical model characterization of the coils, as well as mutual inductance. A capacitor matching network is then determined in order to optimize two objective functions: efficiency of the whole link and power transferred to the load. Since in a practical application the values of the capacitors are discrete and limited to a commercial series of values, we developed a numerical procedure to conduct a search of the best results of both objective functions. This network is characterized by being made up of a series-parallel network at both the input and output. The sequence of this work includes the development of an oscillator used as the primary voltage source and the experimental evaluation of the entire system.
R. Azambuja, V. J. Brusamarello, S. Haffner, and R. W. Porto. “Full Four Capacitor Circuit Compensation for Inductive Power Transfer.” In Instrumentation and Measurement Technology Conference (I2MTC), 2013 IEEE International, 183–187. IEEE, 2013. doi:10.1109/I2MTC.2013.6555406.
Abstract: A novel full four capacitor compensation method for inductive power transfer is introduced. To compute the capacitors values, a very simple search algorithm based on Monte Carlo is used. In addiction, some heuristic are used to reduce the size of the search space. The efficiency, output power and power efficiency are compared with some classical approach such as the two capacitors compensation and also with the basic circuit without compensation. The results showed a significant improvement on the efficiency and output power.
R. W. Porto, V. J. Brusamarello, R. Azambuja, and O. Frison. “Design and Analysis of a GMR Eddy Current Probe for NDT.” In Sensing Technology (ICST), 2013 Seventh International Conference on, 424–429. IEEE, 2013. doi:10.1109/ICSensT.2013.6727688.
Abstract: Defect detection in metallic plates represents an important issue in metal industry, because its potential use in quality control process. Eddy current testing is one of the most extensively used nondestructive techniques for inspecting electrically conductive materials. The purpose of this paper is to present an eddy current testing system for surface defect detection in conducting materials using a giant magnetoresistive (GMR) sensor. An alternate magnetic field is produced by a solenoid and eddy currents are generated in the material under test. The GMR sensor was mounted inside the coil and the arrangement was adapted in the axis of a vertical machining center. In order to validating the measurement device, defects were induced by cracks machined in workpieces made of aluminum. Thus, the parts were scanned with the sensor prototype and a method to estimate the width and depth of the induced defects was proposed after analyzing the output voltage signal.
V. J. Brusamarello, Y. B. Blauth, R. Azambuja, and I. Muller. “A Study on Inductive Power Transfer with Wireless Tuning.” In Instrumentation and Measurement Technology Conference (I2MTC), 2012 IEEE International, 1098–1103. IEEE, 2012. doi:10.1109/I2MTC.2012.6229372.
Abstract: This paper presents the analysis of two loosely coupled coils used to transfer energy to charge a battery. This battery is used to power an electronic device designed to monitor variables such as impact strength, range of temperature and humidity associated with the transport of fruits. The device is inside a sealed enclosure that cannot be opened for recharging the battery. The study shows the loosely coupled coils need to work with a resonance capacitor, at least on the secondary coil. However the resonance frequency also depends on the coupling factor k and the power delivered to the load. Therefore, this work proposes a monitoring system with closed loop for fine-tuning the resonance frequency of the secondary coil circuit. Before starting charging the battery the system scans the resonance frequency on the primary coil and measures the output power on the secondary coil looking for the optimal point. This procedure reduces problems of variation of coupling factor with positioning of the coils.
D. A. Sala, V. J. Brusamarello, R. de Azambuja and A. Cangelosi “Positioning Control on a Collaborative Robot by Sensor Fusion with Liquid State Machines.” In Instrumentation and Measurement Technology Conference (I2MTC), 2017 IEEE International. Milano, 2017. doi:10.1109/I2MTC.2017.7969728.
Abstract: A positioning controller based on Spiking Neural Networks for sensor fusion suitable to run on a neuromorphic computer is presented in this work. The proposed framework uses the paradigm of reservoir computing to control the collaborative robot BAXTER. The system was designed to work in parallel with Liquid State Machines that performs trajectories in 2D closed shapes. In order to keep a felt pen touching a drawing surface, data from sensors of force and distance are fed to the controller. The system was trained using data from a Proportional Integral Derivative controller, merging the data from both sensors. The results show that the LSM can learn the behavior of a PID controller on different situations
R. de Azambuja, F.B. Klein, S.V. Adams, M.F. Stoelen and A. Cangelosi. “Short-Term Plasticity in a Liquid State Machine Biomimetic Robot Arm Controller.” In 2017 International Joint Conference on Neural Networks (IJCNN). Anchorage, 2017 doi:10.1109/IJCNN.2017.7966283.
Abstract: Biological neural networks are able to control limbs in different scenarios, with high precision and robustness. As neural networks in living beings communicate through spikes, modern neuromorphic systems try to mimic them making use of spike-based neuron models. Liquid State Machines (LSM), a special type of Reservoir Computing system made of spiking units, when it was first introduced, had plasticity on an external layer and also through Short-Term Plasticity (STP) within the reservoir itself. However, most neuromorphic hardware currently available does not implement both Short-Term Depression and Facilitation and some of them don’t support STP at all. In this work, we test the impact of STP in an experimental way using a 2 degrees of freedom simulated robotic arm controlled by an LSM. Four trajectories are learned and their reproduction analysed with Dynamic Time Warping accumulated cost as the benchmark. The results from two different set-ups showed the use of STP in the reservoir was useful for one out of three tested trajectories, though not computationally cost-effective for this particular robotic task.
R. de Azambuja, D.H. García, M.F. Stoelen and A. Cangelosi. “Neurorobotic Simulations on the Degradation of Multiple Column Liquid State Machines.” In 2017 International Joint Conference on Neural Networks (IJCNN). Anchorage, 2017. doi:10.1109/IJCNN.2017.7965834.
Abstract: Two different configurations of Liquid State Machine (LSM), a special type of Reservoir Computing with internal nodes modelled as spiking neurons, implementing multiple columns (Modular and Monolithic approaches) are tested against the decimation of neurons, connections and entire columns in order to verify which one can better withstand the damage. Based on the neurorobotics outlook, this work is part of a bigger project that aims to apply artificial neural networks to the control of humanoid robots. Therefore, as a benchmark, we made use of a robotic task where an LSM is trained to generate the joint angles needed to command a simulated version of the collaborative robot BAXTER to draw a square on top of a table. The final drawn shape is analysed through Dynamical Time Warping to generate a cost value based on how close the produced drawing is to the original shape. Our results show both approaches, Modular and Monolithic, had a similar behaviour, however the Modular was better at withstanding the decimation of neurons when it was concentrated in a single column.
Stoelen, Martin F., Ricardo de Azambuja, and Angelo Cangelosi. “A Physical Architecture for Studying Embodiment and Compliance: The GummiArm.” Cognitive Robot Architectures (2017): 68.
Abstract: High bandwidth contacts and unforeseen deviations from planned actions are common in early human development. We here present the GummiArm, an open-source robot with
characteristics that make it interesting for studying development, human motor control, and real-world applications that require robustness and safety. Joints with antagonist actuators
and rubbery tendons provide passive compliance, where the stiffness can be adjusted in real-time through cocontraction. The robot structure is made printable on low-cost 3D
printers, enabling researchers to quickly fix and improve broken parts. The arm has 7+3 Degrees of Freedom (DOF), of which 8 have variable stiffness. It is currently being replicated
across 3 research groups, and we hope to establish a thriving and productive community around this replicable platform.
Abdulla Mohamed, Phil F. Culverhouse, Ricardo De Azambuja, Angelo Cangelosi, Chenguang Yang. “Automating Active Stereo Vision Calibration Process with Cobots.” In IFAC-PapersOnLine
Volume 50, Issue 2, December 2017. doi:10.1016/j.ifacol.2017.12.030.
Abstract: Collaborative robots help the academia and industry to accelerate the work by introducing a new concept of cooperation between human and robot. In this paper, a calibration process for an active stereo vision rig has been automated to accelerate the task and improve the quality of the calibration. As illustrated in this paper by using Baxter Robot, the calibration process has been done faster by three times in comparison to the manual calibration that depends on the human. The quality of the calibration was improved by 120% when the Baxter robot was used.
De Azambuja, Ricardo. “Action Learning Experiments Using Spiking Neural Networks and Humanoid Robots..” PhD thesis, University of Plymouth, 2018.
Abstract: The way our brain works is still an open question, but one thing seems to be clear: biological neural systems are computationally powerful, robust and noisy. Natural nervous system are able to control limbs in different scenarios with high precision. As neural networks in living beings communicate through spikes, modern neuromorphic systems try to mimic them by using spike-based neuron models. This thesis is focused on the advancement of neurorobotics or brain inspired robotic arm controllers based on artificial neural network architectures. The architecture chosen to implement those controllers was the spike neuron version of Reservoir Computing framework, called Liquid State Machines. The main goal is to explore the possibility of using brain inspired neural networks to control a robot by demonstration. Moreover, it aims to achieve systems robust to environmental noise and internal structure destruction presenting a graceful degradation. As the validation, a series of action learning experiments are presented where simulated robotic arms are controlled. The investigation starts with a 2 degrees of freedom arm and moves to the research version of the Rethink Robotics Inc. collaborative humanoid robot Baxter. Moreover, a proof-of- concept experiment is also done using the real Baxter robot. The results show Liquid State Machines, when endowed with an extra external feedback loop, can be also employed to control more complex humanoid robotic arms than a simple planar 2 degrees of freedom one. Additionally, the new parallel architecture presented here was capable to withstand noise and internal destruction better than a simple use of multiple columns also presenting a graceful degradation behaviour.
Jaimes, Arturo F., F. R. de Sousa, F. L. Cabrera, Valner Brusamarello, and Ricardo Azambuja. “Scalable Model of Planar Square One-turn Inductors for Wireless Power Transfer Applications.” In 2018 IEEE Wireless Power Transfer Conference (WPTC), pp. 1-4. IEEE, 2018. doi:10.1109/WPT.2018.8639255.
Abstract: This paper describes an scalable Planar-Square One-turn Inductor model valid for estimating the quality factor and the input impedance of the inductor up to its self-resonance frequency (SRF). Therefore it can be used to guide the optimized process of the inductive link and the power stage of the WPT system without full-wave electromagnetic simulation. The proposed scalar model is analytic, its circuit components have physical meaning, and its model parameters can be extracted from empirical or simulated data. This model was assessed by full-wave EM simulation, the results was consistent the values predicted by the lumped model.
Murliky, Lucas, Rodrigo Wolff Porto, Valner João Brusamarello, Fernando Rangel de Sousa, Alicia Trivino-Cabrera, and Ricardo Azambuja. “Robust active tuning for wireless power transfer to support misalignments and variable load.” In 2018 IEEE Wireless Power Transfer Conference (WPTC), pp. 1-4. IEEE, 2018. doi:10.1109/WPT.2018.8639439.
Abstract: Wireless power transfer has been employed in several applications where the use of cables for powering remote devices are not suitable or feasible. However, such devices work in a tuned frequency and thus the transferred power is set to make the system operate under resonant conditions. These conditions and, in turn the transferred power, are highly dependent on the parameters of an equivalent electrical circuit. Such power can drop abruptly with variations of the relative position between the transmitter and receiver coils as well as the variations of the output load. This work presents a multi-variable with method for maximizing the output power in a wireless power transfer system to make it robust to misalignments and load variations. The described method monitors the magnitude and phase of the input current of the inductive link and proposes an automatic control algorithm by controlling both, the source frequency and a variable matching network. Preliminary results showed that the power delivered to the load can be hold approximately constant over a large range of magnetic coupling coefficient and the load values. These results are a significant improvement of the results obtained by controlling only the source frequency or only the matching network.
Ramtoula, Benjamin, Ricardo de Azambuja, and Giovanni Beltrame. “Data-Efficient Decentralized Place Recognition with 3D Constellations of Objects.” In 2019 International Symposium on Multi-Robot and Multi-Agent Systems (MRS), pp. 219-221. IEEE, 2019. doi:10.1109/MRS.2019.8901073.
Abstract: During the place recognition and optimization steps of multi-robot Simultaneous Localization and Mapping (SLAM), robots need to consider information gathered from other robots working in parallel to build a global map. For these steps, major adaptations of single robot SLAM techniques are necessary to cope with challenges and constraints inherent to multi-robot systems.
B. Ramtoula, R. de Azambuja and G. Beltrame, “CAPRICORN: Communication Aware Place Recognition using Interpretable Constellations of Objects in Robot Networks,” 2020 IEEE International Conference on Robotics and Automation (ICRA), Paris, France, 2020, pp. 8761-8768, doi:10.1109/ICRA40945.2020.9197270.
Abstract: Using multiple robots for exploring and mapping environments can provide improved robustness and performance, but it can be difficult to implement. In particular, limited communication bandwidth is a considerable constraint when a robot needs to determine if it has visited a location that was previously explored by another robot, as it requires for robots to share descriptions of places they have visited. One way to compress this description is to use constellations, groups of 3D points that correspond to the estimate of a set of relative object positions. Constellations maintain the same pattern from different viewpoints and can be robust to illumination changes or dynamic elements. We present a method to extract from these constellations compact spatial and semantic descriptors of the objects in a scene. We use this representation in a 2-step decentralized loop closure verification: first, we distribute the compact semantic descriptors to determine which other robots might have seen scenes with similar objects; then we query matching robots with the full constellation to validate the match using geometric information. The proposed method requires less memory, is more interpretable than global image descriptors, and could be useful for other tasks and interactions with the environment. We validate our system’s performance on a TUM RGB-D SLAM sequence and show its benefits in terms of bandwidth requirements.
Sperling, Moritz, Yann Bouteiller, Ricardo de Azambuja, and Giovanni Beltrame. “Domain Generalization via Optical Flow: Training a CNN in a Low-Quality Simulation to Detect Obstacles in the Real World.” In 2020 17th Conference on Computer and Robot Vision (CRV). IEEE, 2020. doi:10.1109/CRV50864.2020.00024.
Abstract: Many applications in robotics and autonomous systems benefit from machine learning applied to computer vision, but often the acquisition and preparation of data for training is complex and time-consuming. Simulation can significantly reduce the effort and potential risk of data collection, thereby allowing faster prototyping. However, the ability of a data-driven system to generalize from simulated data to the real world is far from obvious and often leading to inconsistent real-world results. This paper demonstrates that some properties of optical flow can be exploited to address this generalization problem. In this work, we train a neural network to detect collisions with simulated optical flow data. Our network, FlowDroNet, is able to correctly predict up to 89 percent of the collisions of a realworld dataset and easily achieves a higher detection accuracy when compared to a network trained on a similar dataset of realworld collisions. We release our code, models and a real-world dataset for collision avoidance as open-source. We also explore the relationship between the complexity of the input information and the ability to generalize to unseen environments, and show that in some situations, optical flow is an interesting tool to bridge the reality gap.