An Adaptive Imitation Learning Framework for Robotic Complex Contact-Rich Insertion Tasks

Complex contact-rich insertion is a ubiquitous robotic manipulation skill and usually involves nonlinear and low-clearance insertion trajectories as well as varying force requirements. A hybrid trajectory and force learning framework can be utilized to generate high-quality trajectories by imitation learning and find suitable force control policies efficiently by reinforcement learning. However, with the mentioned approach, many human demonstrations are necessary to learn several tasks even when those tasks require topologically similar trajectories. Therefore, to reduce human repetitive teaching efforts for new tasks, we present an adaptive imitation framework for robot manipulation. The main contribution of this work is the development of a framework that introduces dynamic movement primitives into a hybrid trajectory and force learning framework to learn a specific class of complex contact-rich insertion tasks based on the trajectory profile of a single task instance belonging to the task class. Through experimental evaluations, we validate that the proposed framework is sample efficient, safer, and generalizes better at learning complex contact-rich insertion tasks on both simulation environments and on real hardware.

Expedited Tensor Program Compilation Based on LightGBM

If the traditional deep learning framework needs to support a new operator, it usually needs to be highly optimized by experts or hardware vendors to be usable in practice, which is inefficient. The deep learning compiler has proved to be an effective solution to this problem, but it still suffers from unbearably long overall optimization time. In this paper, aiming at the XGBoost cost model in Ansor, we train a cost model based on LightGBM algorithm, which accelerates the optimization time without compromising the accuracy. Experimentation with real hardware shows that our algorithm provides 1.8× speed up in optimization over XGBoost, while also improving inference time of the deep networks by 6.1 %.

Design and Prototyping Single-Phase Inverter with Arduino Nano

<p class="Abstract"><span lang="EN-US">Across the year, the needs of Indonesians in the use of electronic equipment are increasing, which results in higher electricity usage. Because most of the electricity load uses AC power, in the application of a DC power source such as solar cells, an inverter that converts DC to AC power is needed. Therefore, the inverter is one of the tools that are widely developed in power electronics. The output voltage from simulation and real hardware is a sine wave with some distortion due to lack of filter; therefore, there occurs a harmonic. The voltage and frequency were also measured with a multimeter. The result shows that both voltage and frequency are closed to the design specification which is 220V 50Hz with the voltage and frequency difference of 1.09% and 0.4%, respectively.</span></p>

On the Energy Performance of Iridium Satellite IoT Technology

Most Internet of Things (IoT) communication technologies rely on terrestrial network infrastructure. When such infrastructure is not available or does not provide sufficient coverage, satellite communication offers an alternative IoT connectivity solution. Satellite-enabled IoT devices are typically powered by a limited energy source. However, as of this writing, and to our best knowledge, the energy performance of satellite IoT technology has not been investigated. In this paper, we model and evaluate the energy performance of Iridium satellite technology for IoT devices. Our work is based on real hardware measurements. We provide average current consumption, device lifetime, and energy cost of data delivery results as a function of different parameters. Results show, among others, that an Iridium-enabled IoT device, running on a 2400 mAh battery and sending a 100-byte message every 100 min, may achieve a lifetime of 0.95 years. However, Iridium device energy performance decreases significantly with message rate.

CCSDS 131.2-B-1 Transmitter Design on FPGA with Adaptive Coding and Modulation Schemes for Satellite Communications

Satellite communications are a well-established research area in which the main innovation of last decade has been the use of multi-carrier modulations and more robust channel coding techniques. However, in recent years, novel advanced signal processing has started being developed for these communications due to the increase in the signal processing capacity of transmitters and receivers. Although signal processing capabilities are increasing, they are still constrained by large limitations because these techniques need to be implemented in real hardware, thus making complexity a matter of critical importance. Therefore, this paper presents the design and implementation of a transmitter with adaptable coding and modulation on a field-programmable-gate-array (FPGA). The main motivation came from the standard CCSDS 131.2-B-1 which recommends that such a novel transmitter which has to date not been implemented in a real system The system was modeled by MATLAB with the purpose of being programmed in VHDL following the AXI-stream protocol between components. Behavioral simulation results were obtained in VIVADO and compared with MATLAB for verification purposes. The transmitter logical circuit was synthesized in a FPGA Zynq Ultrascale RFSoC ZU28DR, showing low resource consumption and correct functioning, leading us to conclude that the deployment of new communication systems in state-of-the-art hardware in satellite communications is justified.

On assessment and evaluation of teaching computer networks to electrical engineering students by the aid of a lab course

In the era of Internet-of-everything, learning the principles of data communications and networking is inevitable for many electrical engineering disciplines. The paper addresses the effectiveness of teaching the fundamentals of data communications and networking using a dedicated lab course as a co-requisite to a classic lecture-based course. In the introduced lab course, the students are asked to do a variety of tasks using real hardware and a network simulator. The paper introduces quantitative measures of an outcome-based learning approach applied to both courses. Based on students’ achievements, the role of the lab course in the attainment of both the course learning outcomes and the electrical engineering program learning outcomes is measured in comparison with the case where the lab course is not taken. Our findings reveal a general enhancement trend in the attainment of the course and program learning outcomes with a significant increase in the program outcome related to solving engineering problems. Also, a slight increase is noticed in meeting the lab course outcomes for the students who attended the lab with the course in the same semester, which indicates an improvement in gaining practical knowledge.

Autonomous Mission of Multi-UAV for Optimal Area Coverage

This study proposes an entire hardware and software architecture from operator input to motor command for the autonomous area coverage mission using multiple unmanned aerial vehicles. Despite the rapid growth of commercial drone services, there are many limitations on operations, such as a low decision-making autonomy and the need for experienced operators to intervene in the whole process. For performing the area coverage mission more efficiently and autonomously, this study newly designs an optimization problem that allocates waypoints created to cover that area to unmanned aerial vehicles. With an optimized list of waypoints, unmanned aerial vehicles can fill the given areas with their footprints in a minimal amount of time and do not overlap each other during the mission. In addition, this study performs both various simulations for quantitative analysis and an outdoor experiment through real hardware implementation in order to verify the performance sufficiently. The methodologies developed in this study could be applied to endless applications using unmanned aerial vehicles equipped with mission-specific sensors.

Analyzing the Performance of WBAN Links during Physical Activity Using Real Multi-Band Sensor Nodes

Wireless body area networks (WBANs) present unique challenges due to their specific characteristics of mobility and over-the-body radio propagation. A huge amount of factors—both internal and external to the network—affect WBAN channel conditions, so a reliable and comprehensive theoretical model of these communications is unfeasible and impractical in real scenarios. Thus, an empirical performance analysis of several WBAN channels is presented in this work, based on the receiver signal strength indicator (RSSI) and packet reception rate (PRR) metrics. Four different static and dynamic activities have been characterized: standing, sitting, cycling and walking. This analysis confirms the theoretical notions of path attenuation due to body parts obstructing the signal path, while serving as a benchmark for the design of future algorithms. The experiments have been carried out with real hardware nodes with wireless interfaces in three ISM bands: 433 MHz, 868 MHz and 2.4 GHz, evaluating the effect of the transmit power and node placement for different subjects. In all scenarios, the PRR metric reaches its maximum of 100% for both sub-GHz bands. Finally, our study concludes that the RSSI metric is sufficient to exploit the periodicity of dynamic activities, without the need for any extra hardware resources.

ACIMS: Analog CIM Simulator for DNN Resilience

Analog Computing In Memory (ACIM) combines the advantages of both Compute In Memory (CIM) and analog computing, making it suitable for the design of energy-efficient hardware accelerators for computationally intensive DNN applications. However, their use will introduce hardware faults that decrease the accuracy of DNN. In this work, we take Sandwich-Ram as the real hardware example of ACIM and are the first to propose a fault injection and fault-aware training framework for it, named Analog Computing In Memory Simulator (ACIMS). Using this framework, we can simulate and repair the hardware faults of ACIM. The experimental results show that ACIMS can recover 91.0%, 93.7% and 89.8% of the DNN’s accuracy drop through retraining on the MNIST, SVHN and Cifar-10 datasets, respectively; moreover, their adjusted accuracy can reach 97.0%, 95.3% and 92.4%.

A real-time HIL control system on rotary inverted pendulum hardware platform based on double deep Q-network

For real applications, rotary inverted pendulum systems have been known as the basic model in nonlinear control systems. If researchers have no deep understanding of control, it is difficult to control a rotary inverted pendulum platform using classic control engineering models, as shown in section 2.1. Therefore, without classic control theory, this paper controls the platform by training and testing reinforcement learning algorithm. Many recent achievements in reinforcement learning (RL) have become possible, but there is a lack of research to quickly test high-frequency RL algorithms using real hardware environment. In this paper, we propose a real-time Hardware-in-the-loop (HIL) control system to train and test the deep reinforcement learning algorithm from simulation to real hardware implementation. The Double Deep Q-Network (DDQN) with prioritized experience replay reinforcement learning algorithm, without a deep understanding of classical control engineering, is used to implement the agent. For the real experiment, to swing up the rotary inverted pendulum and make the pendulum smoothly move, we define 21 actions to swing up and balance the pendulum. Comparing Deep Q-Network (DQN), the DDQN with prioritized experience replay algorithm removes the overestimate of Q value and decreases the training time. Finally, this paper shows the experiment results with comparisons of classic control theory and different reinforcement learning algorithms.