Deep reinforcement learning (DRL) and evolution strategies (ESs) have surpassed human-level control in many sequential decision-making problems, yet many open challenges still exist. To get insights into the strengths and weaknesses of DRL versus ESs, an analysis of their respective capabilities and limitations is provided. After presenting their fundamental concepts and algorithms, a comparison is provided on key aspects, such as scalability, exploration, adaptation to dynamic environments, and multiagent learning. Current research challenges are also discussed, including sample efficiency, exploration versus exploitation, dealing with sparse rewards, and learning to plan. Then, the benefits of hybrid algorithms that combine DRL and ESs are highlighted.

In the last decades, there has been a steady adoption of digital online platforms as learning environments applied to all levels of education. This increasing adoption forces a transition in educational resources which has further been accelerated by the recent pandemic, leading to an almost complete online-only learning environment in some cases. The aim of this paper is to outline the methodology involved in setting up a framework for mapping course-specific data based on student activity to standard learning indicators, which will serve as an input to performance prediction algorithms. The process involves systematically surveying, capturing, and categorising the vast range of data available in digital learning platforms. The data are collected from two sample courses and distilled into five dimensions represented by the generic learning indicators: prior knowledge, preparation, participation, interaction, and performance. The data is weighted based on course development and teaching member’s perspectives to account for course-wise variations. The framework established will allow portability of prediction algorithms between courses and provide a means for meaningful and directed learner formative feedback. Two courses, both bachelor-level and worth 5 European Credits (ECs), that use several online learning platforms in their teaching tools have been chosen in this study to explore the nature and range of student interaction data available, accessible, and usable in a course. The first course is Electromagnetics II at Eindhoven University of Technology, and the second course is Electronics at Delft University of Technology. Both Universities are located in the Netherlands. This work is in the scope of a broader study to use such learning indicators with predictive algorithms to provide a prognosis on individual student performance. The findings in this paper will enable the realization of student performance prediction at a very early stage in the course.

Progress in low power miniaturized electronics and wireless technologies have enabled many innovative applications. Of particular interest is the Internet of Things (IoT) that has dominated the world of ICT in the last half a decade enabling many smart systems and applications. At the same time, we also see much enthusiasm with respect to space missions and applications including terrestrial applications, apart from exploring the universe. We believe that innovations in the domain of IoT will significantly influence the space related activities -- both research and development. In this article, we chart out the innovations in space and the vision with respect to embedded and wireless systems for space applications. In particular, we bring in the notion of Sensor Wireless Actuator Networks in Space (SWANS) and Space Pixels to explain IoT in space. We explain with examples what we envision for the next decade and also the challenges therein. We briefly put forth our four major targets in the next five years.

A new circuit of a magnetic (inductively tuned) RL-multivibrator is proposed. The circuit includes the transconductance amplifier with differential input and single output, and the RL-bridge circuit. The amplifier with limited output current is controlled by the bridge. The output current is distributed between the bridge branches. The resistive branch provides positive feedback and sets the threshold voltage. The inductor branch provides the delay in changing the direction of the amplifier output current. The circuit operation is described, the oscillation wave-forms are calculated, and the oscillation period is found. An example with bipolar transconductance amplifier was realized and tested.

Contrary to what happens with LC oscillators, the increase of coupling in cross-coupled relaxation RC-oscillators leads to a lower quadrature error and lower phase-noise. We present a 2.4 GHz CMOS quadrature relaxation oscillator, and show the measurements confirming this property. Increasing the coupling block gain, the oscillator phase-noise is reduced from -97 dBc/Hz to -104 dBc/Hz @ 1 MHz. The quadrature error is reduced from 4.3deg to 0.8deg. These results show that RC quadrature oscillators may be a practical alternative for RF transceiver applications.

In this paper, an IQ oscillator/mixer circuit is presented in which the mixers are an integral part of the oscillator. The oscillator alone has accurate quadrature outputs, even in the presence of component mismatch. This precise quadrature is preserved when mixers are incorporated in the oscillator. An oscillator/mixer core for a carrier frequency of 5 GHz, suitable for a low-IF front-end, was designed and tested. The power consumption is 30 mW from a 2.5 V supply, and a phase error of 1deg was measured

To maximize the bandwidth of dedicated negative feedback amplifiers by passive frequency compensation, the order of the amplifier needs to be known. Here a method is introduced to determine the order of a circuit with negative feedback. It is shown that the slim of poles in the negative feedback loop, i.e. the loop poles, can be used to determine the order of the amplifier. These loop poles, can be found relatively easily from the circuit diagram and thus the order of the circuit is also relatively easily found.

Applying the non-heuristic biasing theory results in circuits which are optimally biased. However, the resulting circuits will contain many floating voltage sources. This paper describes the use of Integer Linear Programming to minimize the number of these sources.

A biasing method is described intended to make automated biasing of at least some classes of analog circuits straightforward. It has been tested for linear amplifiers, though it is not restricted to that class. A systematic way to introduce bias sources in a circuit is discussed. Also methods for reducing the number of bias sources and bias feedback loops are given. Application of the method has shown that at least for the class of amplifiers the theory is well suited for automation.

An integral method to design systems consisting of several mixers and continuous-time filters, e.g. most radio receiver front-ends, with a maximum dynamic range is developed. The dynamic range of the system obtained by this method is higher then can be attained by designing all filters in the system individually. This method is a generalisation of a design method for single continuous-time filters. The applicability of various optimization and design methods, especially the deviations from regular filter design, is discussed.>

A controllable charge pump is presented that is able to accurately generate a voltage larger than the supply voltage. It operates on a 3 V supply. It is intended to be used to accurately bias MOS-transistors in the triode region, for use as tunable resistors in tunable time continuous filters. The charge pump has been realized in a BiCMOS process (f/sub T/=3 GHz).>