Showing papers presented at "International Conference on Event-based Control, Communication, and Signal Processing in 2020"

Multisampling wave union time-to-digital converter

Abstract: A novel time-to-digital converter principle is presented that introduces multisampling capability to the wave union method. The design employs two wave union launchers for both start and stop signals. For test purposes the converter was implemented in Kintex-7 FPGA device (Xilinx). During the initial experimental verification a mean resolution (LSB) of up to 0.7 ps and precision even below 4.7 ps (RMS resolution better than 3.3 ps) were achieved. The design is FPGA logic resource saving comparing to similar solutions but requires much effort during implementation and data encoding.

Embedded Event-based Visual Odometry

Abstract: This paper presents an event-based visual pose estimation algorithm, specifically designed and optimized for embedded robotic platforms. The visual data is provided by a neuromorphic vision sensor. The fully event-based proposed approach is based on Spiking Neural Networks and a modified Hough transform. The method is developed to detect a square visual feature. The multi-thread algorithm is implemented on a Raspberry Pi, the well-known single-board computer used on many embedded platforms, that is connected to a Dynamic Vision Sensor (DVS) through its USB interface. Validation is done on two different experimental platforms and highlights the ability of the odometry algorithm to determine the relative pose of a robot with respect to a square target, in the aim to be integrated in an event-based visual servoing in a future work.

Event-based Neural Network for ECG Classification with Delta Encoding and Early Stopping

Abstract: We present a scalable architecture based on a trained filter bank for input pre-processing and a recurrent neural network (RNN) for the detection of atrial fibrillation in electrocardiogram (ECG) signals, with the focus on enabling a very efficient hardware implementation as application-specific integrated circuit (ASIC). Our already very efficient base architecture is further improved by replacing the RNN with a delta-encoded gated recurrent unit (GRU) and adding a confidence measure (CM) for terminating the computation as early as possible. With these optimizations, we demonstrate a reduction of the processing load of 58 % on an internal dataset while still achieving near state-of-the-art classification results on the Physionet ECG dataset with only 1202 parameters.

Testing of SCA Waveform Digitization ASIC for High-Precision Time Measurement

Abstract: High speed waveform digitization is one of the methods that can be applied to time measurement. In our previous work, a 5-Gsps switched capacitor array (SCA) chip for high-precision time measurement was designed. In this paper, this SCA ASIC is tested, and the effects of mismatches between sampling cells inside the ASIC are analyzed. The test results show that after the calibration and mismatch correction processes, a time resolution of better than 10ps can be successfully achieved.

Successive Approximation Time-to-Digital Converters

Abstract: The successive approximation scheme belongs to fundamental and most successful methods of analog-to-digital conversion that has been implemented commercially for decades and is still used nowadays. Despite of a widespread use of successive approximation ADCs, the binary search scheme in time-to-digital converters (SA-TDCs) is adopted much rarely. The paper provides an overview and comparison of conversion algorithms for SA-TDCs including the monotone, decision-select, and continuous disassembly successive approximations. The architectures, designs, implementations and performance of SA-TDCs including time resolution, conversion range, conversion time, power consumption, and converter nonlinearities are also discussed.

Multi-channel high-linearity time-to-digital converters in 20 nm and 28 nm FPGAs for LiDAR applications

Abstract: This paper proposes a new calibration method, the mixed-binning method, to pursue a TDC with high linearity in field-programmable gate arrays (FPGAs). This method can reduce the nonlinearity caused by large clock skews in FPGAs efficiently. Therefore, a wide dynamic range tapped delay line (TDL) TDC has been developed with maintained linearity. We evaluated this method in Xilinx 20nm UltraScale FPGAs and Xilinx 28nm Virtex-7 FPGAs. Results conduct that this method is perfectly suitable for driverless vehicle applications which require high linearity with an acceptable resolution. The proposed method also has great potentials for multi-channel applications, due to the low logic resource consumption. For a quick proof-of-concept demonstration, an 8-channel solution has also been implemented. It can be further extended to a 64-channel version soon.

Gray-Code TDC with Improved Linearity and Scalability for LiDAR applications

Abstract: This paper presents a TDC architecture based on a gray code oscillator with improved linearity, for FPGA implementations. The proposed architecture introduces manual routing as a method to improve the TDC linearity and precision, by controlling the gray code oscillator Datapath, which also reduces the need for calibration mechanisms. Furthermore, the proposed manual routing procedure improves the performance homogeneity across multiple TDC channels, enabling the use of the same calibration module across multiple channels, if further improved precision is required. The proposed TDC channel uses only 16 FPGA logic resources (considering the Xilinx 7 series platform), making it suitable for applications where a large number of measurement channels are required. To validate the proposed architecture and routing procedure, two channels were integrated with a coarse counter, a FIFO memory and an AXI interface, to assemble the pulse measurement unit. A comparison between the default routing implementation and the proposed manual routing has been performed, shown an improvement of 27% on the overall TDC single-shot precision. The implemented TDC achieved a 380 ps RMS resolution, a maximum DNL of 0.38 LSB and a peak-to-peak INL of 0.69 LSB, corresponding to a 21.7% and 70.4% improvement, respectively, when compared to the default design approach.

Real time ECG R-peak detection by extremum sampling

Abstract: In this paper, we propose and validate a novel self-adaptive real-time event-driven ECG R-peak detection algorithm based on extremum sampling. The criteria for algorithm development were low complexity and energy efficiency facilitating wearable and mobile applications. The proposed algorithm consists of three major steps: event-driven identification of the QRS window, detection of ECG signal extremum, and awaiting the end of the current QRS complex. The threshold adaptation mechanism reduces the algorithm sensitivity to varying ECG amplitude range. The algorithm performance has been examined on MIT-BIH Arrhythmia Database recordings achieving the following results: sensitivity and positive prediction rate equal respectively to 99.55% and 99.88% which effected in total false detection rate equal to 0.58%. These results significantly outweigh the others reported in the area of low power R-peak detection algorithms.

Event-based Signal Processing for Radioisotope Identification

Abstract: this paper identifies the problem of unnecessary high power overhead of the conventional frame-based radioisotope identification process and proposes an event-based signal processing process to address the problem established. It also presents the design flow of the neuromorphic processor.

Event-triggered coordination of omnidirectional robots over Gazebo

Abstract: This work presents a Gazebo Simulation of a Multiagent system conformed by Omnidirectional (3,0) mobile robots. The model consists of integrating a collaborative control that allows agents to achieve consensus and avoid collisions among them and against obstacles. The control strategy for the communication and agreement is event-based to adapt the simulation to the asynchronous behavior of this type of system, while the collision avoidance is in continuous time. The implementation is carried out relying on ROS and MATLAB/Simulink tools to implement the vehicle model and allow the exchange of information between agents. A virtual leader is introduced to provide the agents with a reference and a trajectory to follow.

Tuning Rules for the Design of MISO Reset Control Systems

Abstract: The PI +CI controller is a hybrid extension of the PI controller that is able to overcome fundamental limitations of linear time invariant control. In this work we propose a new control strategy for multiple-input single-output (MISO) systems consisting of PI +CI controllers connected in parallel, and extend some of the results already known in the single-input single-output (SISO) case. Namely, we find tuning rules that guarantee a flat response after the first reset instant in the case of first order systems without delay, and we study the system's sensitivity to sensor noise by the describing function (DF) method. We compare the performance of this system to that of its linear equivalent by means of two examples.

Event-based Quality-of-Service Parameter Estimation of a Wireless Channel

Abstract: This paper proposes an event-based method to estimate the quality-of-service (QoS) parameters of a wireless communication channel between two unmanned aerial vehicles (UAV). A network estimator determines the QoS parameters at an event time by a channel estimation. Between two event times a model of the channel provides an estimate of the current QoS parameters. A two-state Markov model, which is updated at event times, models the properties of the physical communication channel, which are varying with the distance between the UAVs. In contrast to existing literature the QoS parameters of a varying channel between objects are estimated continuously by a channel model and are utilised for an event-based collision avoidance method, which was introduced in previous publications to cope with an unreliable network. A simulation study with two quadrotors shows the accuracy of the estimate and that collisions can only be avoided by considering the channel properties.

An Effective QRS Selection Based on the Level-Crossing Sampling and Activity Selection

Abstract: The Internet of Things (IoT) healthcare framework is a new trend. In this context, biomedical wearable devices are linked to the cloud. This work contributes to the development of efficient Electrocardiogram (ECG) wearables by redesigning their signal processing chain. The emphasis is on developing a system for effective and precise QRS selection. QRS complexes of heartbeats contain most important arrhythmia related information. The proposed system uses an analog band-pass filter to reduce the impact of Baseline Wander (BW) and aliasing. The band-limited signal is then digitized by using a 5-Bit resolution level-crossing A/D converter. Onward, an original activity selection algorithm is used for an effective selection of the QRS complexes. Results show that the proposed solution attains an average QRS detection sensitivity of 98.4% and positive predictive value of 100% while securing on average 4.77-fold and 2.1-fold compression gains respectively in terms of count of data samples and bits over the classical counterparts.

Arbiterless Event-Based Imager Architecture with temporal and spatial redundancies suppression

Abstract: We present a new image sensor architecture that manages spatial and temporal redundancies with an arbiterless readout and a pixel Address-Event Representation (AER) reading system. This frameless image sensor only generates few events over time in order to target an efficient power consumption compared to the commercial CMOS image sensors. Indeed, this image sensor does not generate anymore frames but events only when a change appears in the scene. Moreover, the event throughput depends on the luminance variations of the recorded scene. This means that more activity in the scene will generate more events and vice versa. Collecting events over a period of time will define an image. It is noticeable that, at each instant, the generated events characterize the zone of interest (the active area) of the scene. Consequently, processing such images should require less computing, communication and energy.

Design of the Time Measurement Circuit, Single Data Frame Sorting, and Packaging Logic of a TOF PET System

Abstract: Positron Emission Tomography (PET) is an important research direction in the field of nuclear medicine. We designed the Single Processing Unit (SPU) electronics for a brain PET system, which is based on the Time-of-Flight (TOF) PET structure. The SPU electronics are required to achieve a 5%0 energy measurement precision as well as a 50-ps-FWHM time measurement precision. The readout electronics include analog front end amplification, shaping, A/D conversion, and time measurement circuits, as well as digital signal processing logic based on a Field Programmable Logic Gate Array (FPGA). Besides, to relieve the data processing pressure of the Coincidence Process Unit (CPU), we designed a single data frame sorting algorithm based on a token ring structure and swap sorting method. The algorithm guarantees that the single data frames are arranged according to their time information. A data frame packaging algorithm is also implemented to ensure the data frames with the time information within a predetermined period are packaged into data frame packages. We conducted tests on the electronics, and the results showed successful time measurement performance while the single data frame sorting algorithm and the data packaging logic functioned well.

Alias-free Discrete-time FIR System Realisation Using Hybrid Stratified Sampling

Abstract: This paper proposes a method for system realisation, where the realised system is described by a continuous-time, finite-duration impulse response. The proposed discrete-time implementation deploys Digital Alias-free Signal Processing. It means that despite the use of digital signal processing, the produced results do not suffer from aliasing. However, owing to the use of random sampling, the approach relies on constructing a suitable estimator of the system output. This paper shows that the proposed estimator is unbiased. It is also consistent, i.e. its variance goes to zero when the density of signal samples increasing. It is proven that under moderately restrictive assumptions, the estimator goes to zero proportionally to the fifth power of the average distance between the samples.

On Random Sampling and Fourier Transform Estimation in Sea Waves Prediction

Abstract: Improving the safety of a wide range of launch and recovery operations is of great international maritime interest. Deterministic sea wave prediction (DSWP) is a relatively new branch of science that can offer such opportunities by predicting the actual shape of the sea surface and its evolution for short time in the future. Fourier transform technique is the main building block in DSWP, which requires measurements of the sea surface. Nonetheless, uniformly sampled measurements of the sea surface cannot be practically achieved for various reasons. Conventional X-band radars are the most realistic candidate to provide a low-cost convenient source of two-dimensional wave profile information for DSWP purposes. Ship movement and mechanically rotating scanning antennas are among sources of irregularity in sea surface sampling. This in turn introduces errors when traditional Fourier transform based wave prediction methods are used. In this paper we show that by modelling the radar sampling instants as random variables and using the estimator of Tarczynski and Allay to process the samples, a reliable solution for DSWP can be constituted.

Distance-Based Send-on-Delta to test the Accuracy of GNSS receivers

Abstract: GNSS accuracy is an important parameter to track vehicle movements in the Earth surface. This paper investigates a new way to dynamically determine the GNSS accuracy based on the Linear-Predicted Send-on-Delta strategy. The proposed method calculates the average deviation of the position measured by a GNSS sensor for a vehicle that moves linearly with constant velocity. Results show that the measured accuracy, named in this work as position-tolerance, is much more precise than the accuracy given by the GNSS sensors, and could be used to improve new position tracking methods.

Prognosis of Health-Related Unavailability of Professional German Football Players

Abstract: In this work, we statistically analyze publicly available data on the health-related absenteeism of German professional football players and, outgoing from this, develop a machine-learning model for a prognosis of the probability that a player will have a longer health problem during the current season. That prognosis is essential for German insurance companies, since a short illness or injury is covered by the health insurance, whereas a longer absenteeism (over six weeks) is covered by a special insurance. Insurance companies have to calculate the risk for that special case in order offer an appropriate insurance rate. Our model gives an assessment of the risk, and thus plays a vital role in the calculation of the insurance rate that can be offered to a professional football player.

Discrete and Continuous Optimal Control for Energy Minimization in Real-Time Systems

Abstract: This paper presents a discrete time Markov Decision Process (MDP) to compute the optimal speed scaling policy to minimize the energy consumption of a single processor executing a finite set of jobs with real-time constraints. We further show that the optimal solution is the same when speed change decisions are taken at arrival times of the jobs as well as when decisions are taken in continuous time.

Tool for Diagnosis of Stochastic Automata

Abstract: This paper is devoted to diagnosis of discrete event systems modeled as stochastic automata. Basic notions of stochastic diagnosis and the functionality of a tool that solves stochastic diagnosers design are introduced.