Showing papers presented at "International Conference on Event-based Control, Communication, and Signal Processing in 2016"

Feature detection and tracking with the dynamic and active-pixel vision sensor (DAVIS)

Abstract: Because standard cameras sample the scene at constant time intervals, they do not provide any information in the blind time between subsequent frames. However, for many high-speed robotic and vision applications, it is crucial to provide high-frequency measurement updates also during this blind time. This can be achieved using a novel vision sensor, called DAVIS, which combines a standard camera and an asynchronous event-based sensor in the same pixel array. The DAVIS encodes the visual content between two subsequent frames by an asynchronous stream of events that convey pixel-level brightness changes at microsecond resolution. We present the first algorithm to detect and track visual features using both the frames and the event data provided by the DAVIS. Features are first detected in the grayscale frames and then tracked asynchronously in the blind time between frames using the stream of events. To best take into account the hybrid characteristics of the DAVIS, features are built based on large, spatial contrast variations (i.e., visual edges), which are the source of most of the events generated by the sensor. An event-based algorithm is further presented to track the features using an iterative, geometric registration approach. The performance of the proposed method is evaluated on real data acquired by the DAVIS.

Steering a predator robot using a mixed frame/event-driven convolutional neural network

Abstract: This paper describes the application of a Convolutional Neural Network (CNN) in the context of a predator/prey scenario. The CNN is trained and run on data from a Dynamic and Active Pixel Sensor (DAVIS) mounted on a Summit XL robot (the predator), which follows another one (the prey). The CNN is driven by both conventional image frames and dynamic vision sensor “frames” that consist of a constant number of DAVIS ON and OFF events. The network is thus “data driven” at a sample rate proportional to the scene activity, so the effective sample rate varies from 15 Hz to 240 Hz depending on the robot speeds. The network generates four outputs: steer right, left, center and non-visible. After off-line training on labeled data, the network is imported on the on-board Summit XL robot which runs jAER and receives steering directions in real time. Successful results on closed-loop trials, with accuracies up to 87% or 92% (depending on evaluation criteria) are reported. Although the proposed approach discards the precise DAVIS event timing, it offers the significant advantage of compatibility with conventional deep learning technology without giving up the advantage of data-driven computing.

ELiSeD — An event-based line segment detector

Abstract: Event-based temporal contrast vision sensors such as the Dynamic Vison Sensor (DVS) have advantages such as high dynamic range, low latency, and low power consumption. Instead of frames, these sensors produce a stream of events that encode discrete amounts of temporal contrast. Surfaces and objects with sufficient spatial contrast trigger events if they are moving relative to the sensor, which thus performs inherent edge detection. These sensors are well-suited for motion capture, but so far suitable event-based, low-level features that allow assigning events to spatial structures have been lacking. A general solution of the so-called event correspondence problem, i.e. inferring which events are caused by the motion of the same spatial feature, would allow applying these sensors in a multitude of tasks such as visual odometry or structure from motion. The proposed Event-based Line Segment Detector (ELiSeD) is a step towards solving this problem by parameterizing the event stream as a set of line segments. The event stream which is used to update these low-level features is continuous in time and has a high temporal resolution; this allows capturing even fast motions without the requirement to solve the conventional frame-to-frame motion correspondence problem. The ELiSeD feature detector and tracker runs in real-time on a laptop computer at image speeds of up to 1300 pix/s and can continuously track rotations of up to 720 deg/s. The algorithm is open-sourced in the jAER project.

An error measurement for resampled level crossing signal

Abstract: This work is a contribution to enhance the signal processing chain required in mobile systems. The system must be efficient in terms of resources utilisation like memory, computation complexity, data transmission, battery consumption, etc. In the aim to achieve these features an event driven signal acquisition, processing and transmission chain based on level crossing and activity selection is adopted. It delivers non-uniformly spaced samples. The employment of efficient and mature classical signal processing tools requires the uniform resampling of level crossing sampled signal. In this context, three resampling techniques, Sample&Hold, Nearest Neighbor and Linear Interpolation, for level crossing sampled signal are examined. Their performance is studied in terms of computational complexity and the resampling accuracy. The resampling error is measured in time domain. It has advantages compared to its counter frequency domain methods. Firstly it does not require the computationally complex frequency domain signal transformation. Secondly it does not introduce distortions, could appear during this frequency domain transformation.

An event-driven building energy management system enabling active demand side management

Abstract: Effective Building Energy Management Systems (BEMS) are essential for future smart grids with implemented demand-response concept. Building Automation and Control Systems (BACS) can gather information about power and energy consumption as well as control loads in buildings and dynamically interact with the smart grid. A demand value is crucial in organizing this kind of the BEMS with an active demand side management (DSM). In this paper we propose different approaches to calculate this demand value using time-driven and event-driven mechanisms characteristic for the BACS. These approaches have been implemented. Experiment with real data has been performed to verify this implementation. Results of experiment show that various demand value computing methods could provide different information about demand level from the active DSM point of view. These approaches have been analysed with respect to the accuracy and speed of computing the demand value. Pros and cons predisposing them to various applications has been found and presented for both the time-driven and even-driven mechanisms.

Event-based state estimation using an improved stochastic send-on-delta sampling scheme

Abstract: Event-based sensing and communication holds the promise of lower resource utilization and/or better performance for remote state estimation applications found in e.g. networked control systems. Recently, stochastic event-triggering rules have been proposed as a means to avoid the complexity of the problem that normally arises in event-based estimator design. By using a scaled Gaussian function in the stochastic triggering scheme, the optimal remote state estimator becomes a linear Kalman filter with a case dependent measurement update. In this paper we propose a modified version of the stochastic send-on-delta triggering rule. The idea is to use a very simple predictor in the sensor, which allows the communication rate to be reduced while preserving estimation performance compared to regular stochastic send-on-delta sampling. We derive the optimal mean-square error estimator for the new scheme and present upper and lower bounds on the error covariance. The proposed scheme is evaluated in numerical examples, where it compares favorably to previous stochastic sampling approaches, and is shown to preserve estimation performance well even at large reductions in communication rate.

Stochastic event-triggered particle filtering for state estimation

Abstract: In this paper, the problem of event-triggered (ET) state estimation is studied for nonlinear non-Gaussian systems. Particle filtering (PF) state estimation approach is developed for systems with stochastic ET measurements to overcome the computational problem in minimum mean square error (MMSE) estimators in which the posterior probability function is non-Gaussian due to ET measurement information. The proposed event triggered particle filtering (ETPF) not only solves the problem of non-Gaussianity but also can handle any functional nonlinearity in the system. It is proved that particles are weighted by the predicted event-triggering (ET) probability density function in the estimator side. The application of the proposed methodology to an interconnected four-tank system is also provided to illustrate and demonstrate the effectiveness of our proposed design methodology.

Levels, peaks, slopes… which sampling for which purpose?

Abstract: There are many ways to nonuniformly sample a signal, the most widely used, but not only, way to perform it being level crossing. The targeted application, and also the characteristics of the signal, can lead to various choices of samplings, to ensure a good representation of the input signal or robustness with respect to noise.

Experimental validation of an event-triggered policy for remote sensing and control with performance guarantees

Abstract: There has been a surge of interest in event-triggered control in recent years, and many event-triggered control methods are now available in the literature. As the theory matures, there is a need to experimentally validate and test these methods in applications of interest. In this paper, we extend and experimentally validate an event-triggered control strategy presented in [1] for the remote point stabilization problem for a ground robot. This strategy specifies when transmissions should occur in both sensor-controller and controller-actuator channels, and guarantees a bound on performance measured by a finite-horizon quadratic cost. The experimental results are coherent with the simulation results and reveal that event-triggered control leads to a tremendous data transmission reduction (up to 90%) with respect to period control, with a minor performance loss.

An event-driven low power electronics for loads metering and control in smart buildings

Abstract: The Internet of Things market is experiencing a period of intense growth. From 7 to 10 billion of connected devices in 2013, it is expected to reach 26 to 30 billion in 2020. Smart buildings represent between 2 to 21% of this market. Indeed, the possibility of connecting objects changes the paradigms of measurement, monitoring and management within the building. This paper presents a miniaturized event-driven low power electronics device dedicated to electrical loads in the context of energy management and optimization systems. An electronic device for distributed intelligence is proposed. It is designed and intended to be added in the path of the power supply of existing electrical appliances. This device helps, as well, to perform local processing of measured data in order to provide a dedicated and optimized control of the connected load.

A 20Mevps/32Mev event-based USB framework for neuromorphic systems debugging

Abstract: Neuromorphic systems are engineering solutions that take inspiration from biological neural systems. They use spike-or event-based representation and codification of the information. This codification allows performing complex computations, filters, classifications and learning in a pseudo-simultaneous way. Small incremental processing is done per event, which shows useful results with very low latencies. Therefore, developing this kind of systems requires the use of specialized tools for debugging and testing those flows of events. This paper presents a set of logic implementations for FPGA that assists on the development of event-based systems and their debugging. Address-Event-Representation (AER) is a communication protocol for transferring events/spikes between bio-inspired chips/systems. Real-time monitoring and sequencing, logging and playing back long sequences of events/spikes to and from memory; and several merging and splitting ports are the main requirements when developing these systems. These functionalities and implementations are explained and tested in this work. The logic has been evaluated in an Opal-Kelly XEM6010 acting as a daughter board for the AER-Node platform. It has a peak rate of 20Mevps when logging and a total of 32Mev of logging capacity on DDR when debugging an AER system in the AER-Node or a set of them connected in daisy chain.

Implementation of a demand elasticity model in the building energy management system

Abstract: Nowadays, crucial part of modern Building Automation and Control Systems (BACS) is electric energy management. An active demand side management is very important feature of a Building Energy Management Systems (BEMS) integrated within the BACS. Since demand value changes in time and depends on various events, factors and parameters, a demand elasticity model has been proposed to provide reliable information about current and expected energy demand. In this paper we propose extension of this model with respect to parameters available in the BACS, determining energy demand level. Real data from the BACS had been imported into a calculation algorithm and proposed approach has been verified in simulation. For easy implementation of the demand elasticity model in the BACS, an extension for logical interface with a new functional profile has been proposed and described. It is ready for integration within the BACS with Internet of Things paradigm.

Event-based control strategy for consensus of a group of VTOL-UAVs

Abstract: This paper presents the development of a collaborative event-based control applied to the problem of consensus and formation of a group of VTOL-UAVs (Vertical Take-off and Landing, Unmanned Aerial Vehicles). Each VTOL-UAV decides, based on the difference of its current state (linear position and velocity) and its latest broadcast state, when it has to send a new value to its neighbors. The asymptotic convergence to average consensus or desired formation is depicted via numerical simulations.

Mobile robot guidance using adaptive event-based pose estimation and camera sensor

Abstract: We present the application of an event-based state estimator to the guidance of a mobile robot. The control feedback loop is closed by means on an estimator based on an Unscented Kalman Filter (UKF) with event-based updates. The estimator requests measurements from a camera sensor only when it needs them. The measurements are triggered by an adaptive condition on the pose estimation error covariance matrix. The main advantage of this method is that it allows for reducing the volume of CPU intensive image processing which is characteristic of these kind of sensors, as well as avoiding the unnecessary overload of the communication channel. The main contribution of the current work is the actual implementation with a P3-DX robotic unit remotely controlled. We discuss practical aspects of the implementation and provide performance results for different parameter settings.

Event-triggered simultaneous fault detection and consensus control for linear multi-agent systems

Abstract: This paper investigates the design and development of event-triggered simultaneous fault detection and consensus control (SFDCC) problem for a network of linear agents. Each agent is equipped with a distributed SFDCC unit that uses the information from the relative measurement sensors and data transmission facilities. It is assumed that the data is transmitted from an agent to its neighbours whenever a specific event condition is satisfied. With the proposed methodology each agent can detect not only its own fault but also the faults occurred in its neighbors and at the same time, all the agents can track the output of a reference model. The parameters of the SFDCC module are designed such that a mixed H ∞ /H − performance index is satisfied based on sufficient conditions in terms of extended linear matrix inequalities (LMIs). It is shown that by using event-triggered data transmission, the amount of data transmitted from an agent to its neighbors and the actuator updates are reduced. Simulation results are presented for fault detection and control of a group of unmanned underwater vehicles to demonstrate and verify the effectiveness of the proposed methodology.

Stability analysis of spatially invariant systems with event-triggered communication

Abstract: In this paper we analyze spatially invariant inter-connections consisting of a (finite) number of subsystems that use packet-based communication networks for the exchange of information. An example of such an interconnected system is the platoon of vehicles that uses cooperative control to drive autonomously. By building upon a recently developed hybrid systems framework for the considered spatially invariant inter-connections, we investigate the design of event-based triggering mechanisms that aim to reduce the amount of transmission times while still guaranteeing behavior in terms of uniform global asymptotic stability (UGAS) for the overall interconnected system. To obtain tractable design conditions, we exploit the spatially invariant property. As a result, we obtain conditions based on only the local information of one of the subsystems in the interconnection and the interconnection structure itself. A nonlinear example is used to illustrate the applications and benefits of the obtained modeling approach.

Event-triggered fault estimation and accommodation design for linear systems

Abstract: The problem of event-triggered active fault-tolerant control (E-AFTC) of discrete-time linear systems is addressed in this paper by using an integrated design of event-triggered fault/state estimator with a fault-tolerant controller. An event-triggered observer is proposed which can simultaneously provide an estimate of the system states and faults. Through an event-triggered transmission mechanism it is shown that the amount of data sent to the observer module is significantly reduced. Moreover, an observer-based fault-tolerant controller based on fault and state estimates is obtained. A robust H ∞ formulation of the problem is given that guarantees stability of the resulting closed-loop system, attenuates the effects of external disturbances and compensates the effects of the fault. Linear matrix inequality (LMI) sufficient conditions are derived to simultaneously obtain the observer and controller parameters and the event-triggered condition. Simulation results for an autonomous remotely operated vehicle (ROV) is given to illustrate the effectiveness of the proposed design methodology.

Optimized design of successive approximation time-to-digital converter with single set of delay lines

Abstract: The paper addresses the problems of design of picosecond resolution time-to-digital converter based on successive approximation (SA-TDC). The principle of the conversion process in SA-TDC consists in successive delaying the events defining a start and a stop of the input time interval by the use of binary-weighted delays. The paper is focused on optimization of particular components of the SA-TDC architecture with a single set of delay lines in order to reduce differential (DNL) and integral (INL) nonlinearities. In particular, the paper contribution is an improvement of time resolution of the converter from 25 ps to 12.5 ps (i.e., by one extra bit) in 180 nm CMOS technology through enhancements of design of circuit components which results in a reduction of conversion errors.

High-level synthesis for event-based systems

Abstract: This paper envisions a design flow for empowering designers in the fast development of low-power event-driven processing chains. This flow takes advantage of level-crossing sampling schemes and asynchronous circuitry. Event-driven paradigm allows better-than-worst-case performance during periods of high-activity of the captured signal as well as a natural stand-by during low-activity periods. The proposed flow uses the specific knowledge of the targeted application and its signals, and a high-level description of the processing algorithm to synthesize a dedicated analog-to-digital converter, which performs the level-crossing sampling, and a digital signal processing unit. The latter is synthesized thanks to a high-level synthesis algorithm following a control/datapath decomposition style. The asynchronous control part is based on distributed asynchronous controllers while the datapath remains similar to a synchronous datapath.

Throughput analysis for error reduction in send-on-delta sampling strategies

Abstract: This paper uses send-on-delta (SoD) strategies to reduce the sampling error instead of reducing the amount of sent messages. For this purpose, we present a procedure that selects an appropriate delta value for a given mean transmission rate. Signals are then sampled using the send-on-delta strategy with the selected delta value. The amount of sent messages will be higher in time intervals where the signal changes fast, and lower in time intervals where the signal changes slow, but maintaining the mean transmission rate and reducing, if possible, the sampling error. The procedure has been implemented for two error-based send-on-delta strategies, and tested on three step response signals. Results show that the proposed procedure decreases significantly the sampling error compared to a conformant periodic transmission strategy.

Event-based visual guidance inspired by honeybees in a 3D tapered tunnel

Abstract: In view of neuro-ethological findings on honeybees and our previously developed vision-based autopilot, in-silico experiments were performed in which a “simulated bee” was make to travel along a doubly tapering tunnel including for the first time event-based controllers. The “simulated bee” was equipped with: • a minimalistic compound eye comprising 10 local motion sensors measuring the optic flow magnitude • two optic flow regulators updating the control signals whenever specific optic flow criteria changed • and three event-based controllers taking into account the error signals, each one in charge of its own translational dynamics. A MORSE/Blender based simulator-engine delivered what each of 20 “simulated photoreceptors” saw in the tunnel lined with high resolution natural 2D images. The “simulated bee” managed to travel safely along the doubly tapering tunnel without requiring any speed or distance measurements, using only a Gibsonian point of view, by: • concomitantly adjusting the side thrust, vertical lift and forward thrust whenever a change was detected on the optic flow-based signal errors • avoiding collisions with the surface of the doubly tapering tunnel and decreasing or increasing its speed, depending on the clutter rate perceived by motion sensors.

Event-based control for a greenhouse irrigation system

Abstract: This work presents a simulation study of an event-based predictive control system for a greenhouse irrigation process. The control system objective is to maintain the desired humidity level, keeping the water usage as low as possible. The event-based control scheme uses a crop transpiration model and a water content virtual sensor to trigger the irrigation system events. In such a scheme, the event-based controller determines the volume of water required to compensate for the irrigation system disturbances. Simulation experiments were performed to analyze the behavior of the designed system and to study the water supply dynamics to the substrate and subsequent drainage and evaporation. The resulting control system is able to adapt the actuation rate to the state of the plant providing the efficient way of water consumption. The obtained results show that application of proposed event-based approach for the greenhouse irrigation system allows us to improve the control performance and to reduce the water usage being an important issue in intensive agriculture. The improved control performance is obtained due to event-based approach and the inclusion of information about the plant dynamic response for water supply and transpiration effect.

Vision-based detection of events using line-scan camera

Abstract: Vision-based method of event detection for control algorithm of acorn scarification device has been presented in the paper. Instead of frame-based approach, line-scan mode of digital camera has been used. The advantage of this approach is an improved temporal resolution of machine-vision-based event detector and reduced latency as the line-acquisition time is significantly lower than the acquisition time of a frame of 2D image. Segmentation of objects moving in front of the camera is one of the steps of the detection. Usage of two alternative methods of segmentation have been reported in the article: Sum of Absolute Differences and motion detection by means of adaptive Gaussian Mixture Model. Motion signal is then a subject of event detection.

A fully-digital and fully synthetizable TDC for high energy physics experiments

Abstract: Trigger and data acquisition systems (TDAQ) of High Energy Physics (HEP) experiments intensively use time measurements for calibration of signals and synchronization between their different elements. Typically, electronics systems for time measurement are designed using a classical mixed-signal approach, while all-digital architectures are nowadays being explored and studied by state-of-art research. Indeed, although an optimized mixed-signal design reaches better performances, it requires a significant amount of design time w.r.t. a fully-digital design. Moreover, an analog IP blocks cannot be easily ported into a new technology, if compared with full digital IPs. In this work, we present a fully-digital TDC application, which is based on a fully synthesizable DCO. The TDC measures the phase relationship between a 40 MHz reference clock and a timing signal under measurement. The DCO design is independent from the technology and it can be described by using a high level Hardware Description Language; moreover, due to its specific characteristics, it can be synthesized, placed and routed by using automatic tools. In this work, we present the architecture of the DCO and of the TDC, their performances and the results on a preliminary implementation on a 130 nm ASIC prototype. The TDC is planned to be used in the upgrade of the Muon detector readout electronic in LHCb experiment at CERN. The TDC presented in this paper has the task of measuring the phase difference between the 40 MHz LHC machine clock and a digital signal coming from the muon detector. One of the main constraints of the TDC is on the phase difference resolution, which has to be about 1.5 ns, in order to cope with to the required resolution of the experiment.

Event-based control as an enabler for high power density processors

Abstract: This paper is part of a long-term research on the application of event-based control to the thermal management of high-power, high-density microprocessors. Specifically, in this work we introduce and discuss a purpose-specific event-based realisation of a digital controller for thermal management, which integrates into a scheme that also takes care of the power/performance tradeoff, and carry out a stability and a preliminary performance analysis. We also present, extending previous research, a comprehensive Modelica library suitable to carry out control studies in the addressed domain. We finally show experimental results on a modern processor architecture, that also compare our solution to the state of the art, to demonstrate the effectiveness of the proposal.

Dynamic event-triggered minimal-order observer for linear systems

Abstract: Event-triggered control is a useful approach in networked control to achieve reduction of data transmission between a plant and a controller. In many practical cases, we can not measure all states of the plant. Then, an observer is often used to estimate the states. On the other hand, the dynamic event triggering mechanism that can make inter-event times larger than those of the conventional event triggering mechanisms has been proposed. In this paper, we consider event-triggered output-feedback control of linear systems. First, we propose an event-triggered minimal-order observer to estimate the unmeasured states from the measured states and control input so that we reduce the computational effort to reconstruct the state. Next, we introduce two dynamic event triggering mechanisms that are extensions of the previous work. One is used at a sensor for the sampling of the output of the plant. The other is at the controller for the update of the control input. We show that the proposed control method ensures the asymptotic stability of the closed-loop system. In addition, we derive triggering conditions under the existence of network delays. Finally, we perform a simulation to show the stability of the closed-loop system and the improvement of the average of the inter-event times by the dynamic event triggering mechanisms.

Event-triggered fault detection for discrete-time LPV systems

Abstract: In this paper, a new event-triggering mechanism is proposed for the problem of fault detection (FD) in discrete-time linear parameter-varying (LPV) systems. A parameter-dependent event-based observer is designed as the residual generator, that only uses the sensor and scheduling variables data that are transmitted only when they are needed. Toward this goal, based on the concept of input-to-state stability, a new formulation is presented to satisfy the H − performance measures and for each performance index, sufficient conditions are given in terms of linear matrix inequalities problems. It is shown that through two event-triggered data transmission mechanisms, the amount of data sent to the fault detection module is decreased significantly. Simulation results demonstrate the effectiveness of the proposed design methodology.

LQG-optimal PI and PID control as benchmarks for event-based control

Abstract: We formulate two simple benchmark problems for event-based control, where the optimal solutions in the continuous-time setting turn out to be ordinary PI and PID controllers. The benchmarks can be used to compare the performance of continuous-time, discrete-time, and various event-based controllers with regard to for instance disturbance attenuation, control effort, and average sampling or actuation rates. They can also be used to evaluate heuristic event-based PI(D) controllers and see how their performance compare to each other and to regular sampled-data control. We give two benchmark examples, where we study the trade-off between event frequency and regulator performance for a number of previously proposed approaches to event-based control.

Periodic event-triggered dual-rate control for a networked control system

Abstract: In this work, a Periodic Event-Triggered Sampling (PETS) strategy is implemented in a Networked Control System (NCS), which uses a gain-scheduled dual-rate PD controller in order to deal with time-varying network-induced delays and packet disorder. Compared to the Time-Triggered Sampling (TTS) strategy, this control solution is able to reduce network utilization (number of transmissions), while still guaranteeing control performance. Simulation and experimental results of an Unmanned Aerial Vehicle (UAV) illustrate the main benefits of the approach.

Building virtualization engine: A novel approach based on discrete event simulation

Abstract: This paper proposes and validates a Discrete Event Simulation (DES) engine for the Smart-Building (SB). It strives to virtualize the common elements found within it and integrates them transparently to an existing Building Management System (BMS), along with existing infrastructure. Thanks to this integration layer, the building management and its control intelligence are completely agnostic to the operations of that virtualization engine. A unique feature of this engine is its micro-treading based core. The latter permits a highly optimized, pseudo-concurrent simulation of hundreds building elements (e.g. loads, sensors, storage, generation, user, etc) in a lean, commodity hardware. Primary virtualization aim is the real-time power data of a SB for the purpose of energy management and Smart Grid connectivity. Moreover, energy optimization and financial studies can be conducted proactively, before committing to costly physical infrastructure investments.