Teaching Control Using NI Starter Kit Robot

Transcription

1 UKACC International Conference on Control 2012 Cariff, UK, 3-5 September 2012 Teaching Control Using NI Starter Kit Robot Payman Shakouri, Member IEEE, (Research stuent) Gorana Collier, Member IEEE, Anrzej Orys, Senior Member IEEE School of Mechanical an Automotive Engineering Kingston University Lonon, UK, SW15 3DW Abstract-Teaching engineering concepts using emonstrations an experiments on real harware is always engaging an well receive by stuents. This paper provies reference materials (both theoretical an test results), to be use in Control teaching an assessment using a laboratory experiment, with a realtime single boar computer base robotic vehicle (National Instruments Robotics Starter Kit). This robotic vehicle is programme using a graphical programming environment. The Aaptive Cruise Control (ACC) algorithm base on Proportional-Integral (PI) an Proportional Integral Derivative (PID) Control are eploye on a fiel programmable gate array (FPGA), inclue in the robot s architecture. The robot moel (base on a given secon orer transfer function) is controlle using the same metho. The results obtaine are compare for the simulation moel an a real robot. The performance comparison emonstrates a goo correlation between theory an implementation, whilst emonstrating problems an iscrepancies introuce by a real system. Keywors: System moelling, Aaptive Cruise Control (ACC), PID Controller, Eucational robotics, Real-time implementation, Single boar computer, FPGA programming. I. INTRODUCTION Lately, Engineering at Kingston University Lonon has been running a project to evelop new Electronics an Control teaching methos. As a result, simulation is not the tool of choice for this task any more. The applie approach to teaching an learning, using latest technology is gaining popularity an interest in the faculty. This paper proposes an experimental metho for teaching control concepts to senior unergrauate engineering stuents an then assessing the learning by setting a group assignment with the objective to esign an implement the software to control the performance of a robotic vehicle with on boar Aaptive Cruise Control (ACC), implemente using a PI an a PID loop. Graphical programming language (LabVIEW) was use for coe implementation, as it oes not eter from the main learning goal by enabling a quick completion of the programming task, allowing stuents to focus on tuning an testing. The target harware was a Starter Kit Robot from National Instruments; where two robots were use as the leaer an the follower. ACC [1], [2] was use to maintain a set istance between them. These experiments not only illustrate the theoretical concepts but also offer an excellent opportunity for stuents to programme a real-time, single boar computer by irectly accessing on-boar FPGA, which is likely to open new opportunities for final year project implementation. The teaching methoology, overall teaching plan (incluing the formal lectures on control esign for istance tracking base on the classical controllers [3], hans-on exercises, an associate assessment task) are presente in Section II. The rest of the paper provies all of the information require by instructors for course elivery, incluing the solution of the assessment, as follows. Section III escribes the harware an associate system transfer function (given to stuents), while Section IV eals with the theoretical concepts require. Section V eals with system test, offering all of the results require by the instructor. Section VI iscusses the esign, tuning an testing of a performance of classical controllers, implemente in ACC. Overall, it verifies the correlation between theoretical expectations an real-life system performance. II. UNDERLYING TEACHING STRATEGY This course is aime at senior unergrauates, who have move on from the system approach use in first year to more iscrete learning. In year two they stuy particular concepts, in orer to acquire sufficient knowlege in a range of isciplines (programming, electronics an control). They can then amalgamate these concepts an work for the first time with a complex system, applying this knowlege an reenforcing unerstaning through implementation. Stuents are given a set of formal lectures, covering relevant control concepts as shown in Table I: Partly sponsore by National Instruments /12/$ IEEE 1037

2 Week TABLE I LECTURE SCHEDULE AND MILESTONES Milestones 1 Introuction to moelling of ynamic systems 2 Use of Laplace transform an transfer functions 3 Time an feeback omain analysis 4 Frequency omain analysis 5 Introuction to PI an PID control incluing PID loop tuning 6 ACC an switching rules This is followe by a set of tutorials an laboratory exercises as shown in Table II. Week 1 TABLE II LABORATORY SCHEDULE, MILESTONES AND MODE Milestones Introuction to the software an harware is one iniviually to ensure that every stuent gets the chance to learn the software an become equippe to work inepenently. However collaborative learning use in later weeks is highly beneficial when new complex concepts are to be aopte. Stuent assessment task aims to: Moe Introuction to control concepts Iniviual 2 PID tuning using Matlab 3 Introuction to LabVIEW an FPGA toolbox 4 Introuction to harware, walkthrough exercises (NI single boar RIO incluing FPGA moule) 5-6 Implementation an testing (work on group assignment) Iniviual Iniviual Group (3) Group (3) 1. Design an Aaptive cruise control algorithm using MATLAB/SIMULINK using the ynamic moel of the robot (given) an obtain optimise solution using PI an PID strategy, in orer to perform the leaer an the follower action (maintain a certain istance between the leaer an follower). 2. Investigate the construction an operation of the real robot, incluing the incorporate sensors an actuators. 3. Investigate a programming metho for a real-time embee FPGA base system. 4. Apply the Aaptive cruise control strategy evelope in 1, to a real robot. 5. Compare system performance in orer to iscuss the correlation between theory an implementation in terms of PID tuning. 6. Compare the system performance to iscuss the correlation between the performance of the robot moel an a real robot. Discuss the problems an iscrepancies introuce by a real system. In aition to the written report incluing the graphs (similar to the ones shown in Sections IV an V), a emo an a viva are also introuce as part of the assessment, to ientify the level of learning for iniviuals an curb plagiarism. Aitional benefits of this assessment task are evelopment of soft skills incluing teamwork, project management an public speaking, with the ae benefit of peer support. III. HARDWARE CHARACTERISTICS AND SYSTEM TRANSFER FUNCTION The LabVIEW starter kit robot (Fig. 1) is base on a NI single- boar RIO, with Ethernet connectivity. The robot follower uses ultrasonic sensors for velocity an istance tracking of the leaer (2cm - 3m sensor range) an two 12V DC brushe motors, (linear velocity of approximately 0.9 [m/s]). The robots can be programme either by using the high-level LabVIEW robotic starter kit API or by using the LabVIEW FPGA moule [4]. The robot moel transfer function approximate through the moel ientification process can be presente as: v vsp 1 = s + s + 1 where v an v sp respectively enote the robot s velocity an velocity setpoint. Units are in [m/s]. (1) 1038

3 Fig.1. NI LabVIEW starter kit robot [4] IV. CONTROL DESIGN The Aaptive Cruise Control is base on two controllers; Proportional-Integral-Derivative (PID) control for velocity tracking, an PI control for istance tracking. The istance tracking controller calculates the velocity setpoint using the ifference of the current inter-istance between the leaing an following robots from the esire istance-heaway. The velocity tracking controller computes the velocity setpoint, to maintain the constant esire cruising spee. ACC concept will be iscusse in the next section, after tuning of the PID parameters. A. PID parameters The PID controller efine in the FPGA moule has the configuration illustrate in Fig. 2. The controller ajusts the spee of the DC motors (in the robot) base on the given set- point spee. The gains of the PID control are set to zero by efault; therefore the set-point spee will be irectly sent to the DC motor. In this work, we esign our controller to calculate the set-point spee which can fulfil the requirement of the ACC system. Different tuning rules to obtain the Proportional-Integralwere introuce in literature Derivative (PID) control gains [5]; such as, step response (frequency response metho), Ziegler-Nichols etc. Control an Estimation Tools Manager toolbox in MATLAB/SIMULINK [10] is use. to tune the PI/PID controllers. By using this toolbox, stuents can easily esign the controllers through the interactive plots such as root-locus, Boe, or Nichols, within the SISO esign tool. The controller can be graphically tune by manually moving, aing, or eleting poles an zeros of the moel s tunable blocks, an observing the close loop response of the system in the analysis plots such as, step response, impulse response, pole/zero response etc. Root locus an open-loop boe iagrams resulte from the tuning process are illustrate in figure 3. These results were obtaine using the Control an Estimation toolbox for a single-input-single- frequency base tuning output system an selecting singular for the tuning algorithm. PID controller gains for the three controllers, use for an ACC system, are given in Table III. (a) Fig. 2. Configuration of the controller incluing PID control in the FPFA Fig. 3. (a) Root locus for open loop system, Open-loop boe iagram 1039

4 TABLE III DISCRETE TIME PID CONTROLLERS GAINS k p k i k Default PID controller Distance Tracking (ACC) Velocity Tracking (CC) The close-loop responses of the system given by Eq.1 are presente in Fig. 4. The responses of the close-loop system in both continuous-time an iscrete-time have been compare. The control action (velocity setpoint v ) has to be constraine between [0,0.9], reflecting the physical limitation of the system i.e. motor velocity. The results obtaine by saturating the control output have been compare to those without saturation, as illustrate in Fig. 4. Fig. 4. shows that the control action initially violates the upper boun limit if it is not constraine. In most of the system, saturation can egrae the performance, especially when using the PID controller. This is ue to a phenomenon calle integrator winup which causes a large overshoot in response. To solve this, an anti-winup can be use. Since there was not any serious issue in our esign, we i not use anti-winup. sp B. ACC switching rules istance (ACC) an velocity (CC) tracking ACC systems operate in two ifferent moes epening on the situation in the front - istance tracking or velocity tracking. If the ultrasonic sensor of the follower (ACC equippe) etects any obstacle, or a slower moving robot in front, the controller ajusts the velocity to maintain the clearance inter-istance (esire istance). If the interistance measure by the ultrasonic sensor is greater than the esire istance, it will switch to velocity tracking moe, known as cruise control (CC) moe, to track the esire velocity. The switching logic was evise to implement automate switching between the two moes of operation for the ACC system, i.e. velocity tracking (CC) an istance tracking [3], [6], [7]. The switching rules for transition between CC an ACC moes are illustrate in Table IV. TABLE IV LOGICAL RULE FOR SWITCHING BETWEEN CC AND ACC (Shakouri, Orys an Laila, et al. 2011) v f < v es v f v es < es ACC CC es CC CC A schematic block iagram of this switching algorithm is unerline in close-up of the switching logic CC/ACC in Fig. 5. (a) Fig.4 (a) the close-loop responses of the system given by Eq.1 using continuous-time an iscrete-time PID controllers, the control actions Fig. 5. A schematic block iagram of an ACC system evelope in the simulation incluing a PI control for istance tracking an a PID control velocity tracking. 1040

5 The esire heaway istance es can be compute using the equation (2), assuming that the follower an leaer robot have the same spee (v f = v l ), known as Constant-Time Heaway policy [8], [9]: es = l + s + Thv 123 f 0 (2) l- robot length, s - aitional istance between two robots, v f - follower robot s velocity an T h is the constant-time heaway (approximates system reaction time) [s]. Distance between real robots is measure by an ultrasonic sensor. In simulation, the istance is etermine by taking the integral of their relative velocity: t = ( v v ) 0 f l t (3) Fig. 7 illustrates the results for istance tracking (ACC) moe for the test execute by using the PID controller. The initial istance between the robots was setup at 0.1 [m]; both robots start from zero velocity. Base on the test, the minimum safe istance o an time-constant heaway T h were chosen to be 0.1 [m] an 1.3 [s], respectively. Some slight unulations in the results on the esire istance curve are ue to the esire istance being a function of the follower robot velocity; therefore its variation epens on the velocity. However, the ACC system operates very well for istance tracking. The Schematic block iagram from Fig 5, implementing ACC in LabVIEW is shown on Fig. 6. a) Fig. 6. The configuration of the LabVIEW block functions for implementing the ACC on a robot V. TESTING AND RESULTS The ACC system was evelope an teste using Matlab for PID tuning an LabVIEW for harware implementation. The programming of the controller has been implemente in LabVIEW MathScript Noe. The ACC system was esigne an teste in the simulation an implemente on the robot s LabVIEW FPGA moule, for comparison. In the initial experiment the leaer robot is virtualise, the sinusoial signal represents the leaer robot s velocity an the istance between robots can be calculate from Eq. 8. The final stage of the test use two test robots; an the istance between them is measure by an ultrasonic sensor. The program evelope in LabVIEW for real-time implementation is epicte in Fig. 6. (c) Fig. 7. Distance tracking (ACC) moe using PI control- (a) The interistance between the robots using test robot, the inter-istance between the robots from simulation, (c) the velocities obtaine uring istance tracking using test robot an from simulation implementation Fig. 8 shows the results for velocity tracking (CC) moe accomplishe by PID Control with the same initial conitions. However, the esire cruising spee has been selecte as a step form; from 0-20 [s], the value of the 1041

6 esire spee is 0.3 [m/s], its value reuces to 0 [m/s] between [s] an then it reverts to 0.3 [m/s] from 40 [s] onwar. Since the actual istance is far away from the esire istance, the ACC system operates in CC moe, so that the follower robot s velocity tracks the esire cruising spee (see Fig. 8.c), rather than tracking the esire istance. Consequently, it causes the istance between the robots to be increase (see Fig. 8.a-b). (a) 1. System moelling an simulation. 2. Implementation in harware. 3. Testing of the ACC system. This paper eals with the implementation of only one type of control algorithm an there is scope to implement others e.g. fuzzy logic, or even Moel Preictive Control, which can be use for post grauate level teaching. There is also potential for system harware enhancement/expansion, to inclue further sensors, for example global positioning systems (GPS), inertial navigation systems (INS), camera etc. (all presently available off the shelf for the specifie target boar), opening a range of aitional opportunities for further control feeback but also experiments an assignments in the fiel of igital imaging, signal processing, system fusion, robotics, autonomous riving an more. The teaching materials escribe in this paper can be mae available on request to the conference participants. REFERENCES (c) Fig. 8. Velocity tracking (CC) using PID control- (a) The inter-istance between the robots using test robot, the inter-istance between the robots from simulation implementation, (c) the velocities obtaine uring istance tracking using test robot an from simulation implementation VI. CONCLUSIONS The experiment escribe here elivers a range of robotics an control concepts, an control strategies, incluing PI an PID, PID tuning an switching strategies. It requires stuents to program these strategies an to implement in a real evice a target real-time microcontroller boar that has been configure as a robot, which is an excellent target system, capturing stuents interest an imagination. This paper proposes an assessment brief for the experiment to contain: [1] Martinez, J. J., an C. Canuas-e-Wit. A Safe Longituinal Control for Aaptive Cruise Control an Stop-an-Go Scenarios. Vol. 15. IEEE Transactions on Control Systems Technology, [2] Xiao, Lingyun, an Feng Gao. A comprehensive review of the evelopment of aaptive cruise control system. Vehicle System Dynamics 48: 10 (April 2010): [3] Shakouri, P., A. Orys, D. S. Laila, an M. R. Askari. Aaptive Cruise Control System: Comparing Gain-Scheuling PI an LQ Controllers. 18th IFAC Worl Congress. Milano, Italy, [4] National Instrument. 15 September [5] Ogata, K. Moern Control Engineering. 3r. The Unite State of America: Prentice-Hall, [6] Shakouri, P., an A. Orys. Application of the State-Depenent Nonlinear Moel Preictive Control In Aaptive Cruise Control System. 14th International IEEE Conference on Intelligent Transportation Systems - ITSC Washington, DC, USA, at The George Washington University, October 5-7, [7] Geres, J C, an J K Herick. Vehicle Spee An Spacing Control Via Coorinate Throttle An Brake Actuation. (Control Eng. Practice) 5 (1997): [8] van en Bleek, R. A. P. M. Design of a Hybri Aaptive Cruise Control Stop-&-Go system. Master s Thesis, TNO Science & Inustry, Technische Universiteit Einhoven, [9] Zhou, W., an S. Zhang. Analysis of Distance Heaways. Proc. Of the Eastern Asia Society for Transportation Stuies [10] Control system Toolbox User's Guie. The MathWorks, Inc.,