Hardware-In-the-Loop (HIL) Testbed for Evaluating Connected Vehicle Applications

Hardware-In-the-Loop (HIL) Testbed for Evaluating Connected Vehicle Applications Department of Mechanical Engineering University of Minnesota Project Members : Mohd Azrin Mohd Zulkefli Pratik Mukherjee Yunli Shao Prof. Zongxuan Sun Date : 10/16/15 (Friday) FPIRC 2015 1

Presentation Outline Introduction Powertrain Research Platform Hardware-In-the-Loop (HIL) Testbed Introduction Test Results with Simple Traffic Network Test Results with Complex Traffic Network Conclusions Future Directions 2

Background Road Site Unit Traffic Center Glossary IVC : Inter Vehicle Communication VII : Vehicle-Infrastructure-Integration DSRC : Dedicated Short Range Communication Intelligent Vehicles Travel direction Detectors IVC & VII are introduced to improve safety and mobility. Information exchange between vehicles are supported by : DSRC communication standards [1] : IEEE 802.11p Wireless Access in Vehicular Environments (WAVE). IEEE 1609 Security, Network Service & Multi Channel Operation. SAE J2735 Message Set Dictionary for Basic Safety Message (BSM). FCC Allocate 5.85 5.925 GHz band for DSRC communication. 4

Motivation Evaluation of connected-vehicle application in real traffic is difficult and time consuming with safety and legal concerns. Inaccurate fuel and emission maps requires the use of real engine. Microscopic traffic simulation can mimic actual traffic if calibrated and driven by real traffic inputs. Previous Methods and Challenges Inaccurate fuel and emission maps in simulations [2]. Difficulties and space requirements to instrument on-road vehicles with big measurement devices [3-4]. Safety and legal concerns to test connected vehicle in real traffic [5]. Proposed Research Previous Methods Deficiencies HIL Testbed Development of HiLS for EMS Evaluation Inaccurate fuel-use and emission maps. Difficulties & space requirements to instrument on-road vehicles. Safety and legal concerns. HiLS measures real engine fuel-use and emissions. Testing done in lab & engine is easily instrumented and replaced. Realistic simulated traffic does not pose safety or legal concerns. 5

Powertrain Research Platform 7

Working Principles Hardware Components Main Dynamics ω e = T e J e T pump = T e J e J e T friction J e D M 2πJ e P out + D M 2πJ e P in T friction J e P out = β e V t2 q in β e V t2 q out β e V t2 = β ed M 2πV t2 q leak ω e β ec d A HS V t2 2 ρ P out w HS Wang, Y., Sun, Z., and Stelson, K.A., Modeling, Control, and Experimental Validation of a Transient Hydrostatic Dynamometer, Control Systems Technology, IEEE Transactions on, v19, n6, pp. 1578-1586, Nov 2011. β e V t2 q leak Valve opening w HS is controlled to track ω e and engine throttle angle is used to control T e 8

Control Architecture Three-level control architecture : 1) High Level : EMS to optimize reference (T e, ω e ). 2) Middle Level : Virtual Powertrain Model calculate desired engine load. 3) Low Level : Dynamometer control - track desired engine load. Wang, Y., Sun, Z., and Stelson, K.A., Nonlinear Tracking Control of a Transient Hydrostatic Dynamometer for Hybrid Powertrain Research, Proceedings of the ASME 2010 Dynamic Systems and Control Conference, pp. 61-68, September 12-15 2010. 9

Overview of Hardware in the Loop System (HiLS) SMART-SIGNAL Field Data Processer Zoom-In intersection Powertrain Research Platform Signal Controller Cabinet Glossary Controller Middleware Middlewares to connect the different HiLS components Physical Components Software/Hardware Components Microscopic Traffic Simulator (VISSIM) Connected Veh Middleware Connected Veh Controller Powertrain Middleware HiLS Component Purpose Ownership Powertrain Research Platform Controls load to real engine for fuel & emission measurements. U of MN Microscopic Traffic Sim (VISSIM) Simulate traffic & provide speed trajectory to Powertrain Res. Platform. BOTH Connected Vehicle Controller Controls vehicles in VISSIM for connected vehicle applications. U of MN SMART-SIGNAL Provide real traffic input to VISSIM simulation. U of MI Signal Controller Cabinet Controls a virtual intersection in VISSIM. U of MI 11

HIL Testbed Powertrain Middleware Powertrain COM Desired-veh-speed is extracted from VISSIM, while actual-veh-speed is calculated from the powertrain dynamics using actual engine speed and torque. Powertrain Research Platform Remote computer running VISSIM Dyno (Hardware) Actual T e, ω e Powertrain Dynamics (Simulation) Desired/ Optimized T e, ω e MATLAB-Simulink Control/ Optimization Control of Powertrain model & Optimization Actual Veh Speed Des Veh Speed Desired Veh Speed VISSIM Input VISSIM COM Desired Veh Speed VISSIM Input VISSIM Traffic Simulator Traffic Solid arrows indicate local communication. Dashed arrows indicate remote communication via C# Socket Programming. 12

HIL Testbed Powertrain Middleware Powertrain COM Currently, one-way communication is implemented for testing before implementing two-way communication. Powertrain Research Platform Remote computer running VISSIM Dyno (Hardware) Desired/ Optimized T e, ω e MATLAB-Simulink Control/ Optimization Desired Veh Speed Desired Veh Speed VISSIM COM Desired Veh Speed VISSIM Traffic Simulator Actual T e, ω e Powertrain Dynamics (Simulation) Control of Powertrain model & Optimization Traffic Solid arrows indicate local communication. Dashed arrows indicate remote communication via C# Socket Programming. 13

Test Setup : Traffic Network 1700 meters long with 7 traffic-lights (fixed-timing) at every 200m between 300m and 1500m. Vehicle speed data was transferred from a computer running VISSIM to powertrain research testbed remotely at every 0.2 seconds. Vehicle with no-stop, 1-stop, 2-stops and 3-stops were identified before tests were conducted for each vehicle. HIL Video 300m 1500m 15

Fuel Use & Emissions Vehicle Dynamics Test 1 : No Stop 16

Fuel Use & Emissions Vehicle Dynamics Test 2 : 1-Stop 17

Fuel Use & Emissions Vehicle Dynamics Test 3 : 2-Stop 18

Fuel Use & Emissions Vehicle Dynamics Test 4 : 3-Stop 19

Grams of diesel fuel Test Results : Fuel Consumption & Emissions 55.26 57.97 76.59 102.3 0.04 0.23 1.01 1.17 2.03 183.42 0.05 0.27 1.28 1.36 2.37 184.79 0.06 0.34 1.60 1.64 2.85 242.81 0.09 0.43 2.09 1.95 3.43 329.75 Fuel Consumption Emissions 120 4 HCHO NO2 CO 400 100 80 60 40 20 Grams of HCHO, NO2, CO, NO and NOX 3.5 3 2.5 2 1.5 1 0.5 NO NOx CO2 350 300 250 200 150 100 50 Grams of CO2 0 0 no-stop 1-stop 2-stops 3-stops 0 20

Test Setup : Traffic Network 3.5km stretch on Medical Drive between Babcock Road & Fredericksburg Road in San Antonio, TX. Traffic Simulation Complexities : Multiple vehicle types : cars, busses & trucks. Multiple lanes with lane-changing. Varying speed limits for roads & lanes. 7 signalized & 6 non-signalized intersections. Reduced vehicle speeds, right-of-ways & pedestrian crossings at intersections. Stop signs at non-signalized intersections. Public transportation stops. Two vehicles with 2-stops & 3-stops traveling the same route are selected for test cases. 22

Fuel Use & Emissions Vehicle Dynamics Test 1 : 2-Stop 23

Fuel Use & Emissions Vehicle Dynamics Test 2 : 3-Stop 24

Test Results : Fuel Consumption & Emissions Grams of diesel fuel 112.2 138.4 0.11 0.56 2.69 2.57 4.51 353.58 0.12 0.69 3.24 3.45 6.00 437.62 Fuel Consumption Emissions 160 140 120 100 80 60 40 20 Grams of HCHO, NO2, CO, NO and NOX 7 6 5 4 3 2 1 HCHO NO2 CO NO NOx CO2 500 450 400 350 300 250 200 150 100 50 Grams of CO2 0 2-stops 3-stops 0 2-stops 3-stops 0 25

Conclusions Vehicle data from remote traffic simulation extracted and transferred in real-time to the powertrain research platform over the internet through COM interfaces and socket programming. Different vehicle speed profiles accurately tracked by powertrain research platform to represent the target vehicle in VISSIM simulation. Simple powertrain optimization employed in powertrain research platform to optimize engine operating points in real-time, which can be extended to complex optimization methods utilizing traffic data in the future. Real fuel and emissions measurements are recorded, which can be used to evaluate optimization methods for connected vehicle applications in the future. 27

Future Directions Upgrade one-directional communication to two-directional to reflect actual vehicle speed from powertrain research platform in VISSIM simulation. Build connected vehicle controller and middleware to process traffic data from VISSIM simulation. Calibrate VISSIM traffic simulation with real-traffic from data collected on instrumented vehicle and highway (cooperation with MnDOT). Support the benefits evaluations of connected vehicle technologies from accurate fuel consumption and emissions measurements on the testbed. Support benefit assessments of several USDOT s connected vehicle applications : Eco-Approach, CACC, Eco-Driving and Speed Harmonization. 29

References 1. Kenney, J.B., Dedicated Short-Range Communications (DSRC) Standards in the United States, Proceedings of the IEEE, v99, n7, pp. 1162-1182, July 2011. 2. Filipi, Z., Fathy, H., Hagena, J., Knafl, A. et al., Engine-in-the-Loop Testing for Evaluating Hybrid Propulsion Concepts and Transient Emissions - HMMWV Case Study, SAE Technical Paper 2006-01-0443, 2006. 3. Duoba, M., Ng, H., and Larsen, R., Characterization and Comparison of Two Hybrid Electric Vehicles (HEVs) - Honda Insight and Toyota Prius, SAE Technical Paper 2001-01-1335, 2001. 4. Hu, H., Zou, Z., and Yang, H., On-board Measurements of City Buses with Hybrid Electric Powertrain, Conventional Diesel and LPG Engines, SAE Technical Paper 2009-01-2719, 2009. 5. Hall, R.W. and Tsao, H.S.J., Automated Highway System Deployment: A Preliminary Assessment of Uncertainties, Automated Highway Systems, pp. 325-334, 1997. 30

Backup Slide 1 : CACC Controller x d = k p (x p x d_a d 0 ) + k d ( x p x d_a ) + x p x d = x d dt x p x p x p x d_a x d_a x d d 0 = Preceding-vehicle speed (from VISSIM) = Preceding-vehicle acceleration (from VISSIM) = Preceding-vehicle distance travelled (from VISSIM) = Actual follower-vehicle speed (from HIL) = Actual follower-vehicle distance travelled (from HIL) = Desired follower-vehicle speed (CACC controller output) = Desired spacing (constant) By choosing appropriate k p and k d gains, the error dynamics will stabilize to zero. Therefore, the distance between preceding and following vehicle can be kept constant. 31

Backup Slide 2 : HIL Testbed with Embedded CACC Controller Powertrain COM Vehicle speed extracted from VISSIM is assumed to be lead vehicle speed CACC controller use this information to calculate follower vehicle speed using fixedspacing car-following policy. Dyno acts as the follower car. Powertrain Research Platform Remote computer running VISSIM Dyno (Hardware) Desired/ Optimized T e, ω e Control/ Optimization Follower Veh Speed CACC Controller Lead Veh Speed Lead Veh Speed VISSIM COM Lead Veh Speed VISSIM Traffic Simulator Actual T e, ω e Powertrain Dynamics (Simulation) Control of Powertrain model & Optimization MATLAB-Simulink 32

Test Results Follower vehicle enters traffic network 20s after lead vehicle enters Follower vehicle catches-up with lead and maintain 3-meters (10 feet) spacing Zoom In between 80s - 200s 33

Emissions Vehicle Dynamics Results with CACC Controller (Different Rule- Based EMS) Fuel or Emission Gas Fuel Consumed (g) NO x (g) NO (g) NO 2 (g) HCHO (g) CO (g) CO 2 (g) Total 58.47 2.7719 1.5775 0.3505 0.0669 1.7031 182.6519 34

Backup 3 : Rule Based Method P wheel = T v ω v PSOC = SOC target SOC K fit Rule-Based Map for John Deere Engine P req = P wheel + P SOC T e = P e ω e = P req ω e Iterate ω e and select minimum m fuel T e, ω e. A Rule-Based map can be iterated offline at different values of P req to create a mapped correlation between minimum m fuel T e, ω e and P req (see Figure above) 35