Deliverable D54.3 Active Green Driving: 1 System Functionality
Transcription
Deliverable D54.3 Active Green Driving: 1 System Functionality
Highly automated vehicles for intelligent transport 7th Framework programme ICT-2007.6.1 ICT for intelligent vehicles and mobility services Grant agreement no.: 212154 The future of driving. Deliverable D54.3 Active Green Driving: st 1 System Functionality Version number Dissemination level Lead contractor Version 0.4 CO, PU Continental Automotive GmbH Due date 31.10.2010 Date of preparation 22.08.2011 HAVEit 22.08.2011 Authors Name Company Martin Sanfridson VTEC Astrid Lundgren VTEC Johan Jarlengrip VTEC Grant Grubb VTEC Lei Feng VTEC Maria Bruce VTEC Project Managers Prof. Dr. Alfred Hoess Continental Automotive GmbH Siemensstrasse 12 93055 Regensburg, Germany Phone +49 941 790-5786 Telefax +49 941 790 99 5786 E-mail: [email protected] Holger Zeng Continental Automotive GmbH Siemensstrasse 12 93055 Regensburg, Germany Phone +49 941 790 92330 Fax. +49 941 790 99 92330 E-mail: [email protected] Project Co-ordinator Dr. Reiner Hoeger Continental Automotive GmbH Siemensstrasse 12 93055 Regensburg, Germany Phone +49 941 790 3673 Fax +49 941 790 13 3673 E-mail [email protected] Copyright: HAVEit Consortium 2010 D54.3 Active Green Driving 1st System Functionality 2 HAVEit 22.08.2011 Revision and History Chart Version Date Reason 0.1 01.10.2010 First version, a compilation and read through (Martin Sanfridson) 0.2 22.10.2010 Revision at internal review (Martin Sanfridson) 0.3 13.11.2010 Final editing and submission 0.4 22.08.2011 Rewording of sentence according to reviewers comments 06.10.2011 Final editing and preparation of final version for submission to EC; generation of reduced deliverable version intended for publication D54.3 Active Green Driving 1st System Functionality 3 HAVEit 22.08.2011 Executive Summary The overall objective of the HAVEit project is to develop technical systems to improve automotive safety and efficiency. In WP5400, Volvo Technology contributes to the overall objective by developing a fuel efficiency focused Active Green Driving (AGD) application. This document reports on the first results of the AGD application. In all parts of the application, simulation on desktop PC and on more elaborated driving simulators with drivers in the loop, have been planned steps in the development. In the AGD application, information obtained by sensors installed on the demonstration vehicle to monitor the external traffic environment, are used to predict the future driving horizon. By predicting near future changes (within the horizon), as for example speed and elevation, the, in the best case optimal, use of the demonstration vehicle‟s hybrid powertrain can be found and executed automatically without intervention from the driver. The key difference to a conventional powertrain is the brake regeneration and temporary storage of electrical energy. Furthermore, without restrictions to the kind of powertrain deployed, information within the prediction horizon is also used in a Driver Coaching System to help the driver handle the vehicle in a more fuel efficient manner by advice via graphical display and haptic accelerator pedal. These are the functionalities that are evaluated in this report. The AGD demonstration vehicle is up and running as a test bench for speed and load prediction and subsequent optimal control of the powertrain and for driver coaching. The prediction is based both on static route information taken from the in-vehicle database, using GPS to acquire the vehicle‟s position such as maximum likely speed and bus stops, but it is also based on dynamic traffic ahead taken from the front mounted laser scanners. The first vehicle tests show that the estimation of future vehicle speed works well. Currently, there is work in progress to make the prediction more robust. The Energy Management works well in simulation and shows that fuel actually can be saved. The functionality is currently deployed in the vehicle, but the installation has not been completely debugged. The Driver Coaching System has been installed in the vehicle, including screens and haptic pedal, and tests finalized within one month. To complement the previous report [6], additional results from the sensor installation and integration will be described. D54.3 Active Green Driving 1st System Functionality 4 HAVEit 22.08.2011 Table of Contents Executive Summary....................................................................................................................................... 4 Table of Contents .......................................................................................................................................... 5 List of Figures ................................................................................................................................................ 6 List of Tables ................................................................................................................................................. 7 1 Introduction ............................................................................................................................................. 8 2 Active Green Driving ............................................................................................................................ 10 2.1 Functionalities ............................................................................................................................... 10 2.1.1 Speed and load prediction ..................................................................................................... 10 2.1.2 Energy Management ............................................................................................................. 13 2.1.3 Driver coaching ...................................................................................................................... 14 2.2 Architecture ................................................................................................................................... 16 2.2.1 Architecture and bus structure ............................................................................................... 16 2.2.2 Adaptation of the generic HAVEit architecture and deployment ........................................... 17 2.2.3 AGD specific architecture modifications ................................................................................ 20 3 Demonstrator configuration .................................................................................................................. 21 3.1 Perception layer: Sensors ............................................................................................................. 21 3.1.1 Laser Scanners ...................................................................................................................... 21 3.1.2 V2V ........................................................................................................................................ 23 3.1.3 Sensor Data Fusion (SDF) .................................................................................................... 26 3.2 Command layer: Joint System ...................................................................................................... 28 3.3 Execution layer: actuators ............................................................................................................. 30 3.4 Driver interface .............................................................................................................................. 32 3.4.1 Driver‟s display....................................................................................................................... 33 3.4.2 Demonstration passengers‟ display ....................................................................................... 36 3.4.3 Haptic gas pedal .................................................................................................................... 37 3.5 System integration ........................................................................................................................ 37 4 Preliminary AGD system validation ...................................................................................................... 39 4.1 Relevant use cases and scenarios ............................................................................................... 39 4.1.1 AGD use case 1: Predicted deceleration and acceleration ................................................... 39 4.1.2 AGD use case 3: Predicted downhill (uphill) driving .............................................................. 40 4.1.3 DCS use case 1: Speeding .................................................................................................... 40 4.1.4 DCS use case 5: DCS Report – Post trip feedback .............................................................. 42 4.2 Scenario 1: AGD use case 1 ......................................................................................................... 43 4.3 Scenario 2: AGD use case 3 ......................................................................................................... 44 4.4 Scenario 3: DCS use case 1 – Speeding ..................................................................................... 45 4.5 Scenario 4: DCS use case 5 – Post trip feedback ........................................................................ 48 5 Conclusions .......................................................................................................................................... 49 References .................................................................................................................................................. 50 Annex 1: Keywords ...................................................................................................................................... 51 D54.3 Active Green Driving 1st System Functionality 5 HAVEit 22.08.2011 List of Figures Figure 1: Sensor location overview of the WP5400 demonstration vehicle .................................................. 9 Figure 2: Schematic flow diagram of the AGD Speed and Load prediction. ............................................... 11 Figure 3: Schematic flow diagram for the Energy Management (using maps). .......................................... 13 Figure 4: Schematic flow diagram for the Driver Coaching System. ........................................................... 15 Figure 5: Network topology.......................................................................................................................... 16 Figure 6: Top view of installed sensors and control units on the demonstrator vehicle .............................. 17 Figure 7: Architecture overview of the HAVEit AGD system ....................................................................... 18 Figure 8: AGD deployment diagram, with functional modules and layers .................................................. 19 Figure 9: Top view laser scanner visualization, with accompanying video ................................................. 22 Figure 10 Object distance measurements for the scene in Figure 9 .......................................................... 23 Figure 11: The vehicles used for V2V system verification .......................................................................... 24 Figure 12: Test route driven during part of the V2V communications testing ............................................. 25 Figure 13: Graphs indicating the transmission of CAN messages transmitted in V2V testing ................... 25 Figure 14: Sensor data fusion architecture ................................................................................................. 26 Figure 15: Example SDF visualization for tracked objects .......................................................................... 27 Figure 16: Speed prediction as a function of distance. Blue line is predicted and red is actual. ................ 29 Figure 17: Speed prediction as a function of time. Blue line is predicted and red is actual. ....................... 29 Figure 18: Torque prediction as a function of distance. .............................................................................. 30 Figure 19: Vehicle speed (left axis, blue line) and gas pedal (right axis, fat green line). ............................ 31 Figure 20: Generated torque by combustion engine (left, yellow) and mass flow of fuel (right, green). ..... 31 Figure 21: Electrical machine torque (left, yellow) and state of charge (right, green). ................................ 32 Figure 22: Examples of GUI‟s of the second version. ................................................................................. 33 Figure 23: Example of the popup strategies. ............................................................................................... 34 Figure 24: Example of the GUI for the post trip feedback screen. .............................................................. 35 Figure 25: Hybrid driveline and prediction horizon view. ............................................................................. 36 Figure 26: Drivers view GUI and prediction horizon view. .......................................................................... 36 Figure 27: Haptic gas pedal installation ...................................................................................................... 37 Figure 28: Predicted deceleration followed by acceleration. ....................................................................... 39 Figure 29: Uphill (downhill) driving. ............................................................................................................. 40 Figure 30: DCS use case 1 actor events ..................................................................................................... 41 Figure 31: DCS use case 5 actor events ..................................................................................................... 42 Figure 32: Energy optimization during a start and stop drive cycle ............................................................ 44 Figure 33: Energy optimization during a single hill drive cycle .................................................................... 45 Figure 34: Driving simulator......................................................................................................................... 46 Figure 35: Driving simulator interior setup ................................................................................................... 46 D54.3 Active Green Driving 1st System Functionality 6 HAVEit 22.08.2011 List of Tables Table 1: Summary of installed sensors and their capabilities ....................................................................... 9 Table 2: Summary of output data from the SDF system ............................................................................. 27 Table 3: DCS user comments ..................................................................................................................... 47 D54.3 Active Green Driving 1st System Functionality 7 HAVEit 1 22.08.2011 Introduction The intention of this document is to report intermediate results of the AGD application and from the demonstration vehicle. After this brief introduction, the functional architecture and deployment structure will be described (section 2). In the following section 3 the turn comes to subsystem verification (i.e. comparing with requirements) and the results from the first field test. In section 4 the intermediate validation will be presented following the sample use cases set up prior to the work. Requirements and specifications for the AGD application and the demonstration vehicle are found in [1] and [2]. Previous documents have reported on sensor installation [5], and the functional architecture [6]. The main goal of the Active Green Driving is to reduce the fuel consumption and potentially emissions in a heavy hybrid demonstrator vehicle by using an e-Horizon system enhanced with environment sensors. This is accomplished by two quite different strategies, both relying on the continuous prediction of vehicle speed: 1. A fuel optimal powertrain strategy, where the overall control objectives are, for any moment in time, to decide: the power-split between the electrical machine and the internal combustion engine, temporary switching on and off the internal combustion engine, and gear shifts by considering powertrain efficiency at selected revolution speed and torque. 2. A driver support interface to give driving recommendations, in certain defined situations, using both graphical and textual information and by haptic feedback in the accelerator pedal. The objective is of course to minimize fuel consumption. None of these strategies offers automated driving – the driver is in charge of acceleration and deceleration in the traditional way. The first strategy simply accepts the driver‟s behaviour and adjusts the powertrain accordingly, and the second strategy tries to influence the driver by giving advice. The two strategies work independent of each other. To accelerate the hybrid vehicle, the driver effectively requests a torque to the wheels. The requested torque has to be realized by the powertrain actuators. In a hybrid vehicle, there is more flexibility, compared to a conventional vehicle, to realize this torque. The demonstrator vehicle has a parallel powertrain which means the power to the wheels can at any moment be a mix of power from the combustion engine and power from the electrical machine. When the driver brakes, kinetic energy is recuperated by operating the electrical machine as a generator and store the electrical energy in the high voltage battery. The e-Horizon system provides information about future driving conditions several minutes ahead of the vehicle. The information from this system is of static nature, which means that it does not change from time to time during the same route. Typical information provided is speed limits of the road and the location of, for instance, traffic lights, roundabouts and altitude. The other three sensor systems, the laser scanner, the camera and the V2V communication, give environmental information about future driving conditions a few seconds ahead of the vehicle. All sensors combined will provide the demonstration vehicle, a Volvo city bus, with information on the long term static road conditions as well as short term dynamic traffic environment in front of the bus. D54.3 Active Green Driving 1st System Functionality 8 HAVEit 22.08.2011 Electronic horizon Color camera IR-V2V communication Laser scanners Figure 1: Sensor location overview of the WP5400 demonstration vehicle The environment sensors in Table 1 are described in detail in [5]; see also Figure 1 for an overview. Sensor Measured Quantities Range Field of View Laser scanners Frontal objects, road data 150 m 150 degrees Colour camera Traffic light data, objects (if possible) 50 m 60 degrees IR-V2V (vehicle to vehicle) communication Data received from the preceding vehicle 15/50 m (two modes used simultaneously) 25/10 degrees (two modes used simultaneously) Electronic horizon Coordinates & map data - - Table 1: Summary of installed sensors and their capabilities D54.3 Active Green Driving 1st System Functionality 9 HAVEit 2 22.08.2011 Active Green Driving In this section we will dive into AGD application to explain how it works internally and how it is structured according to the HAVEit architecture. 2.1 Functionalities The two main functionalities in AGD are Energy management, and Driver coaching. The quality they deliver relies heavily on input from two important auxiliary functions: Sensor fusion, and Speed and load prediction. The speed and load prediction gets its input in an array that represents preview information assembled by sensor fusion of the external sensors, but also some vehicle internal signals, such as the vehicle speed. A good description of sensor fusion can be found in [6]. In this section speed and load prediction, energy management and driver coaching will be treated. These functionalities are executed mainly on the hardware unit called AGD-ECU. For an overview of the hardware installed in the demonstration vehicle for the AGD project see Figure 5. 2.1.1 Speed and load prediction The main functions and interfaces of the speed and load prediction are presented in the signal flow diagram in Figure 2. Different shapes and colours are used in the flow diagram to denote internal process (light blue), external process (dark blue), data signals and database (turquoise). Yellow boxes are used to label the conveyed signals which are symbolized by the arrows. The rough order of processing according to Figure 2 is to first read and handle the external input and to continue with vehicle internal status and driver behaviour. The speed and load prediction takes the vehicle‟s coordinates from the navigation device, based on GPS. To get a more reliable vehicle position filtering is necessary. There is an option, especially in situations with poor reception, to activate a Kalman filter which utilises speed, yaw rate and inclination of the vehicle to calculate an enhanced prediction of the current position, [8]. The map matching translates the position from the coordinate system used by the GPS into a new coordinate system that is more suitable when later extracting data from the eHorizon route database which holds stored routes. First of all the application needs to know which of the stored routes the user has planned to drive. This can be set via the driver‟s HMI or from the AGD-ECU, but it is also possible to search for the closest stored route. It is assumed that a route exists. D54.3 Active Green Driving 1st System Functionality 10 HAVEit 22.08.2011 GPS Signals from Vehicle CAN Vehicle speed Yaw rate Inclination Position coordinates Filtering Map matching eHorizon Route Database Route identifier Route distance Sensor Fusion eHorizon provider Road slope position and angle Speed limit position and value Event position and type Curvature position and magnitude Current vehicle position Preceding vehicle data Preceding fixed object data Traffic light state Signals from Vehicle CAN Vehicle Speed Speed Plan Road slope position (m) and angle (%) Speed position (m) amd Speed plan (m/s) Speed plan limit attribute (numeral) Current vehicle position Signals from Vehicle CAN Vehicle Speed Vehicle acceleration Vehicle mass Speed and Torque prediction with driver model Predicted vehicle speed array Predicted required torque array Energy Management Driver Coaching System Display of prediction and powertrain status Figure 2: Schematic flow diagram of the AGD Speed and Load prediction. D54.3 Active Green Driving 1st System Functionality 11 HAVEit 22.08.2011 When a set of data points within a predefined maximum distance from the current filtered position is found in the database, the search for the closest data point will be performed along those points. The distance from the current position to a point on the route is defined as the orthogonal projection from the current position point to the road. The travelled distance from the start of the route will be the new position coordinate which is the lookup key to the database: Route distance. The eHorizon data function uses this information to extract a horizon of road attributes. The eHorizon route database has the following main attributes: Speed limit position [m] Speed limit [m/s] Number of speed limit pairs Event position [m] Event type o speed bump o give way sign o pedestrian crossing o roundabout o traffic light o bus stop o stop sign Number of event pairs Curve position [m] Curvature value [1/m] Number of curvature pairs Slope position [m] Slope value [ ] Number of slope pairs From the sensor fusion an array of preview information is constructed and continuously updated. This is derived from the laser scanners and the colour camera and consists primarily of the speed of and distance to any vehicle in front, the distance to the nearest fixed object in front and the state of the nearest traffic light ahead. Next, information from the database and from the sensor fusion are combined in the process called speed plan. The output from the speed plan will be an expected upper limit of the vehicle speed over a distance horizon. Thereafter the speed plan is converted to a time vector and driver behaviour is taken into account. The torque necessary to propel the vehicle is also calculated based on known physical properties using Newton‟s second law. The result is later used by the Energy Management and the Driver Coaching System. D54.3 Active Green Driving 1st System Functionality 12 HAVEit 2.1.2 22.08.2011 Energy Management Speed and load prediction Predicted vehicle speed array Predicted required torque array State of Charge Equivalence factor calculation module Equivalence factor Electric Motor torque Diesel engine torque Hybrid powertrain Maps module Suggested gear Gear shift cost Suggested Mode Mode cost Current gear Desired torque Current speed Current Mode Cost evaluation module Requested gear Requested mode Figure 3: Schematic flow diagram for the Energy Management (using maps). An overview of the architecture of the Energy Management, allocated to the AGD-ECU, is shown in the flow diagram of Figure 3. As in the previous section, the blue boxes represent modules / functions and the yellow boxes label the interconnecting signals. The Energy Management receives signals from the speed and load prediction and also communicates with the Hybrid powertrain via Vehicle CAN bus HLP-1 and HLP-7. The Energy management consists of three parts: an Equivalence factor calculation module, a Maps module and a Cost evaluation module. The control signals, i.e. the output from the Energy Management, are: the torque of the electric motor (in units of Nm), the gear selection (number 1 to 12) and the powertrain mode (i.e. electric only, hybrid or Diesel only drives). The demanded torque is satisfied firstly by the electrical motor up to its current limits and secondly by the combustion engine to satisfy the remaining demanded torque, [9]. The currently deployed energy management algorithm in the AGD demonstration vehicle is an equivalent consumption minimization strategy (ECMS), commonly found in the academic literaD54.3 Active Green Driving 1st System Functionality 13 HAVEit 22.08.2011 ture see e.g. [12] that introduces an optimization criterion in order to choose the best combination of control signals. The equivalence factor calculated in the equivalence factor module might be explained as the price of the electricity stored in the battery, compared to the Diesel fuel. There are losses in the transformation of energy that mainly cause heat which means that efficiency is decreased. The high voltage battery allows recuperation of energy originating in the Diesel fuel, but the chain of transformation has a lower efficiency than Diesel only mode. Generally, it pays off to recuperate electric energy by regenerative braking – this is preferably done at low speed. However, since auxiliary equipment is attached to the high voltage system, this is not always sufficient but has to be complemented by regeneration of electric energy when the Diesel engine is leaving additional torque. This is all taken care of in the Energy Management controller. The optimization is computationally intensive. To enable real time execution of the optimization, intermediate results are pre-computed off-line and stored in arrays (look-up tables), handled by the Maps module. The static maps are functions of the desired torque, the current speed of the vehicle, the current gear and the equivalence factor. They are used for the hybrid drive mode. If a shift in gear or powertrain mode is suggested, this is first evaluated in the cost evaluation module to determine which energy gain can be made in pursuing with the shift, before the demand is sent to the hybrid powertrain. 2.1.3 Driver coaching The flow diagram in Figure 4 gives an overview of the functional modules of the Driver Coaching System (DCS). As before, different shapes and colours are used in the diagram to denote internal functional module (light blue), external functional module (dark blue), data signals (yellow). First, the maximum recommended speed is calculated by monitoring the continuously updated incoming speed plan array (see Figure 4), from the AGD Speed and Load Prediction, and the current vehicle speed. The KPI events detection monitors driver actions to detect misbehaviour in one of the Key Performance Index (KPI): Upcoming object, speeding, harsh braking and downhill. The Eco points are, for each KPI, increased and decreased to reflect driver behaviour. There is a module to monitor the KPIs, prioritize and trigger DCS tips and DCS feedback, and control pop-up timing of this information on the driver‟s display. The priority order for the pop-ups is as follows, from highest to lowest priority: 1. Upcoming Object 2. Bad planning 3. Speeding 4. Downhill An overall instant eco value is calculated based on the individual eco points for each KPI; see section 4.1.3 for more details. When the driver stops the vehicle and requests information, the summary module generates a comprehensive summary of eco points and events. The haptic pedal control generates the necessary data to provide tactile feedback in order to instruct the driver on better driving practices. The driver can enable the DCS with a switch (called DCS switch logic in the figure), see [7] for more details. The driver coaching switch has three positions: “off”, “on” and “summary”. The post drive summary can be reset to stop accumulating to previous summary. D54.3 Active Green Driving 1st System Functionality 14 HAVEit 22.08.2011 Speed and load prediction Vehicle CAN Vechicle speed Predicted vehicle speed array Find recommended speed Max recommended speed Vehicle CAN Vechicle speed Brake pedal Acc pedal KPI events detection Haptic pedal KPI Events Generate and display Feedback/Tips Calculate KPI Eco points KPI eco points Compute Eco value Tips/ Feedback Instant Eco value Summary module Driver‟s Display Driver DCS switch logic Figure 4: Schematic flow diagram for the Driver Coaching System. D54.3 Active Green Driving 1st System Functionality 15 HAVEit 22.08.2011 2.2 Architecture This section is devoted to the HAVEit architecture and hardware structure specific for the AGD. 2.2.1 Architecture and bus structure Laser Scanner GPS Receiver Laser Scanner USB Ethernet RS232 Ethernet LS Switch box USB RS232 Front Cam. CAN 3 CAN 1 HMI PC HMI PC CAN 4 Sensor Fusion CAN V2V CAN 3 CAN 1 V2V CAN 2 LS ECU Sensor Fusion xPC ECU (xPC) AGD ECU Auto Box (microautobox) CAN 4 1 Ethernet CAN 1 Ethernet DACU CAN 2 Vehicle CAN HLP1 Vehicle CAN HLP7 Pedal Figure 5: Network topology. Figure 5 shows the hardware structure with processing units and communication channels [11]. The Vehicle CAN bus HLP1 and HLP7 connect the AGD specific units to the already existing ECUs. On these CAN buses the AGD ECU can communicate with the Transmission ECU and the Engine ECU and listen to the messages sent from the rest of the ECUs in the vehicle. The AGD ECU (a Micro Autobox), Sensor Fusion ECU (an xPC) and HMI PC (running a recent version of Windows NT) are able to communicate with each other and exchange information on a private bus in the figure called Sensor fusion CAN. To connect external sensors etc. there are additional Ethernet, CAN, USB and RS232 communication standards. The Sensor fusion ECU is connected to two laser scanners (LS) via a switch and an additional processing ECU, a front looking colour camera (FrontCam), a front vehicle to vehicle (V2V) comD54.3 Active Green Driving 1st System Functionality 16 HAVEit 22.08.2011 munication device, a back V2V communication device and a yaw rate sensor (DACU). The information from the different sensors is processed and passed on to the rest of the AGD system. The yaw rate sensor is needed by the camera to compensate for the yawing motion that represents panning. Besides the communication with the AGD nodes the Sensor Fusion ECU is also able to listen to CAN messages sent on CAN HLP1. The display for the driver is connected to the HMI PC. Engineering stations for development and tuning of the AGD system are primarily connected to the Sensor Fusion ECU and AGD ECU for development purpose. The approximate physical location of the sensors, ECUs and wiring are shown in the top view of the demonstration bus in Figure 6, see [6] and [11]. Existing platform units are not displayed. Ethernet Power supply CAN USB RS232 V2V V2V Driver LS switch LS ECU DACU xPC AGD ECU HMI PC GPS receiver Power station & panels Laser scanner Acc. Pedal Driver‟s display Camera/ECU Laser scanner Figure 6: Top view of installed sensors and control units on the demonstrator vehicle 2.2.2 Adaptation of the generic HAVEit architecture and deployment This section will give an overview of the AGD specific adaptation of the generic HAVEit architecture and its deployment. A diagram of the AGD specific functional architecture (Figure 7) is also found in [6], where it is derived from the generic HAVEit architecture, see the HAVEit specification deliverable [2] or [10] for details. There are some differences due to the fact that in AGD the driver is not monitored to select e.g. an automatic driving mode. The architecture consists of four layers: 1. Perception layer (component C1, colour code blue), 2. Command layer (component C2, colour code yellow), 3. Execution layer (component C3, colour code green), and 4. Driver interface components layer (component C4, colour code red). To avoid confusion, it should be pointed out that the use of the term “layer” in this diagram differs from that in traditional computer engineering terminology. D54.3 Active Green Driving 1st System Functionality 17 HAVEit Driver interface components 22.08.2011 Environment sensors Vehicle sensors Driver Perception layer Sensor data fusion HMI Command layer Co-pilot Prediction system Command generation and plausiblization Vehicle speed and torque demand vector Execution layer Drivetrain control Braking actuator Electric motor actuator Combustion engine actuator Gearbox actuator Figure 7: Architecture overview of the HAVEit AGD system The generic function block diagram, including processing units and communication channels, for the HAVEit architecture allocation to hardware units in [3], with the purpose to find a feasible and practical allocation of tasks and messages, has to a large extent been followed in the AGD version. Figure 8 shows the allocation of the various software modules found in previous figures onto the three main processing units used in AGD: Sensor Fusion ECU, AGD ECU and HMI PC. The layers of the architecture including its layer components have been depicted in the cases a software module is inside any of the three processing units. The figure is focused on showing what is inside the processing units, not the relation between modules or layers for which the other views in Figure 2, Figure 3, Figure 4, and Figure 7 serve the purpose better. The AGD ECU hosts the Energy Management functionality in the Drive train control. The Copilot in the AGD system handles the prediction of the maximum vehicle speed and torque demand trajectories for the nearest future, knowing that the vehicle will drive on a defined route [6]. The prediction in the Command Generation and Plausibilization module handles the dynamic parts in the prediction of the vehicle speed and torque trajectories. The main task for the Command Generation and Plausibilization module is to derive predicted vehicle speed and torque demand trajectories based on vehicle dynamics and a driver model. In this module also driver coaching instructions are derived based on the driver‟s driving behaviour and upcoming traffic situations. The HMI PC handles the eHorizon positioning system and is connected to a GPS receiver, which provides information about the current position of the vehicle. The vehicle position is processed in the HMI PC and then passed on to the AGD ECU, where a map database with GPS positioned corresponding objects are stored. The database information forms the eHorizon system outputs, which are used by the speed and load prediction. The HMI PC also runs the graphics for the driver‟s display. The sensor fusion of external sensors is run on the Sensor Fusion ECU [6]. The eHorizon route data base has an instance in the HMI PC and also in the AGD ECU to minimize the communication need. The HMI-PC supplies information on the active route identifier and position along that route to the AGD-ECU where road attributes are stored locally. D54.3 Active Green Driving 1st System Functionality 18 HAVEit 22.08.2011 Sensor Fusion ECU HMI PC eHorizon Route Database Co-pilot Filtering Sensor data fusion HMI Map matching Sensor Fusion Driver‟s Display AGD ECU Co-pilot eHorizon provider Speed plan Driver interface components Command generation and plausiblization Speed and torque prediction with driver model Find recommended speed Perception layer eHorizon Route Database Command layer Execution layer KPI Events detection Calculate KPI Eco points Drive train control Cost evaluation module Equivalence factor calculation module Maps module Compute eco value Generate and display feedback/tips Summary module Haptic pedal Figure 8: AGD deployment diagram, with functional modules and layers D54.3 Active Green Driving 1st System Functionality 19 HAVEit 2.2.3 22.08.2011 AGD specific architecture modifications To have a common and generic architecture in the HAVEit project is an advantage, but the aims and requirements of the AGD subproject are in the deployment reflected in another set of sensors, actuators and other hardware components necessary for the implementation. These units and components are omitted from the generic architecture [3]: Driver State Assessment is not needed since there are no active automation functions. Active steering is not needed Active braking is not needed Mode Selection and Arbitration Unit (MSU) is not needed. The AGD Driver Coaching System is enabled by the driver with a switch on the dash board. Instead, the focus is on the AGD ECU with its speed and load prediction and Energy Management. D54.3 Active Green Driving 1st System Functionality 20 HAVEit 5 22.08.2011 Conclusions The preliminary validation of the AGD functionality by use of test scenarios relies on the use of simulation. Due to technical problems with the powertrain of the demonstration vehicle, it has so far not been possible to carry out all validation scenarios in the vehicle. However, simulators are often a good substitute for early validation of ideas and for verification of algorithms. A full range of scenarios will be presented in a later report at the end of the project. The Speed and Load prediction, a prerequisite for the AGD functionality of saving fuel, has had a first test in the demonstrator vehicle. The estimated speed profile horizon ahead of the vehicle is very similar to the actual speed profile when comparing the logged data files. It shows that the prediction works as intended. However, the traffic light detection has not been taken in full service yet. New hardware was installed in late September 2010 and a new version of software was delivered first week of October. The vehicle to vehicle infra red communication sensor has been tested but it will not be used by the speed and load prediction algorithm. The Energy Management controller of a hybrid vehicle can be made more fuel efficient by using information about future driving conditions. The value of additional information is however difficult to assess, since the possibility of taking advantage of the information depends on the capability of the vehicle, i.e. constraints in the transformation and storage of energy. This is what the simulation of the demonstrator vehicle above shows. The implemented Energy Management control works as expected in simulation. Testing of the Energy Management in-vehicle has not been done yet. In real world applications, information about future driving conditions needs to be estimated from e.g. map data and measurements from sensors, which will not be fully accurate. It is therefore very important that the Energy Management is robust against errors in the Speed and Load Prediction. The robustness against both estimation errors and different prediction horizons has to be evaluated in the future. Some of the main challenges with developing a Driver Coaching System involve how to present the information to the driver to keep him motivated. The information and advice need to be intuitively understood and easy to ignore by the driver if focus is needed elsewhere. In order to gain acceptance the system also needs to manage the trade-off between fuel-economical driving and performance in such a way that it is possible to follow the recommendations of the system without unreasonably affecting the performance or traffic flow. There are Eco KPIs (Key performance indices) that the driver can affect (e.g. speeding, hard braking) and there are aspects (possibly Eco KPIs) that the driver cannot affect, e.g. vehicle load and route planning. In order to develop a fair Eco driving functionality, e.g. for back-office use, this second type of aspect must also be taken into consideration. The preliminary studies in the driving simulator, although not by professional drivers, show the difficulties in getting immediate acceptance, maybe since it challenges the driver‟s idea of fuel economic driving. The Eco KPIs must be related to certain threshold values and the driver needs to be able to relate his performance to a threshold value. The reference value should preferably be the optimum value, the best value that a driver can score within a certain KPI. An initial set of these tuning parameters can preferably be found in the driving simulator, where the scenario is easy to repeat. The driver‟s HMI, i.e. display of DCS related information and advice and the haptic pedal, has also been tested with good result, but more on-road tests is needed. D54.3 Active Green Driving 1st System Functionality 49 HAVEit 22.08.2011 References [1] HAVEit deliverable D11.1 “Requirements“, September 2008 [2] HAVEit deliverable D11.2 “Specification”, February 2009 [3] HAVEit deliverable D12.1 “Architecture“, February 2009 [4] HAVEit internal document “HAVEit V2V CAN Specification”, August 2010 [5] HAVEit deliverable D54.1 “AGD: Sensors installed in vehicle (1st SW version), May 2009 [6] HAVEit deliverable D54.2 “AGD: Components installed, working and tested”, May 2010 [7] Internal report ER-540261 “Eco Driver Coaching – HAVEit”, work in progress [8] Internal report on Speed and load prediction, work in progress [9] Internal report, “Energy management for the AGD project”, work in progress [10] Internal report ER 59251, “HAVEit WP 5400, Active Green Driving – Software interface and architecture, July 2010. [11] Internal report ER 59252, “HAVEit WP 5400, Active Green Driving – Hardware installations”, 2010. [12] Sciarretta, A. Back, M. Guzzella, L. „Control Strategy of Parallel Hybrid Electric Vehicle„, IEEE Transactions on control systems technology, Vol. 12 No. 3, may 2004, p 352-363 D54.3 Active Green Driving 1st System Functionality 50 HAVEit 22.08.2011 Annex 1: Keywords Abbreviation Description AGD Active Green Driving AQuA Automated Queue Assistance CAN Controller Area Network DACU Gyro sensor ECU at Volvo DCS Driver Coaching System ECMS Equivalent Fuel Consumption Minimization Strategy ECU Electronic Control Unit eHorizon A system where GPS is used in combination with a digital map EM Electrical Machine (high voltage) GPS Global Positioning System HAVEit Highly Automated Vehicles for intelligent transport HLP Higher Level Protocol HMI Human-Machine Interface ICE Internal Combustion Engine KPI Key Performance Index MAB dSpace‟s Micro Autobox for rapid prototyping with special h/w PC Personal Computer SF, SDF Sensor Fusion, Sensor Data Fusion SoC State of Charge (of the high voltage battery) V2V / I2V Vehicle to vehicle communication / Infrastructure to Vehicle WP Work Package xPC Mathwork‟s rapid prototyping using PC hardware D54.3 Active Green Driving 1st System Functionality 51