|
1
|
|
|
2
|
- A way of driving that reduces fuel consumption and greenhouse gas
emissions – Ecodriving.org
- Teaching eco-driving can improve actual fuel efficiency by an average of
17 percent - McKinsey & Company 2009
- 1% of the highway trip is responsible for 16, 19, 4, 3, and 4% of the
trip’s total HC, CO, NOX, CO2, and fuel
consumption – Ahn and Rakha 2008
|
|
3
|
- Develop an Eco-drive system
- Predictive Eco-Cruise Control (ECC) system
- Eco-car-following
|
|
4
|
- Describe the building blocks of the Eco-drive system
- Fuel Consumption Model
- Powertrain Model
- Predictive Eco-Cruise Model
- Car-Following Model
- Overview of proposed algorithm
- Simulation results
- Study conclusions and recommendations
|
|
5
|
|
|
6
|
- Virginia Tech Comprehensive Power-based Fuel consumption Model (VT-CPFM)
- Has the ability to produce a control system that does not result in
bang-bang control and
- Easily calibrated using publicly available data without the need to
gather detailed engine and fuel consumption data.
- Estimates CO2 emissions (R2=95%)
|
|
7
|
|
|
8
|
- Typical powertrain models:
- Computationally intensive
- Challenging to integrate within microscopic traffic simulation
software
- Require proprietary parameters
- Require gathering field data for the entire envelope of operation of a
vehicle.
- Simple vehicle powertrain model needed:
- CSM developed a model used within the context of this approach
|
|
9
|
- The proposed model
- Uses driver throttle input to compute the engine speed and power and
finally compute the vehicle acceleration, speed, and position
- Can be calibrated using vehicle parameters that are publically
available without the need for field data collection.
|
|
10
|
- 2007 Chevy Malibu: I-81 southbound
- 65 mph cruise control operation
- Measured: 13,297 kW vs.
Estimated: 13,871 kW (4.3% Error)
|
|
11
|
|
|
12
|
- The proposed predictive eco-cruise control system
- Generates optimal vehicle controls using topographic data.
- Optimizes the vehicle controls in advance using a dynamic programming
(DP) implementation of Dijkstra’s shortest path algorithm.
- Requires three system parameters:
- Discretization distance (or stage length), the look-ahead distance,
and the optimization frequency.
|
|
13
|
- Three step optimization:
- Define search space using powertrain model
- Speed and gear space that the vehicle is physically able to achieve
for the given topography and vehicle characteristics
- Discretize continuous search space
- Use speed and gear levels to construct a graph
- Compute optimum control (minimum path)
- The vehicle speed and gear changes over each stage considering a cost
function at stage transitions
|
|
14
|
|
|
15
|
|
|
16
|
- Car-following models define relationships between a following and
preceding vehicle in a range of inter-vehicle spacing.
- Modeled as
- Equations of motion under steady-state conditions plus
- Constraints that govern the behavior of vehicles while moving from one
steady-state to another.
- The Rakha-Pasumarthy-Adjerid (RPA) model is used
- Van Aerde steady-state car-following model
- Vehicle dynamics acceleration and deceleration constraints
|
|
17
|
- Steady-State Modeling
- where sn(t) is vehicle spacing at time t, un(t) is
speed of vehicle n at time t (km/h), uf is free-flow speed
(km/h), △t is length of time interval, c1 is fixed
vehicle spacing constant (km), c2 is first variable vehicle
spacing constant (km2/h), and c3 is second
variable vehicle spacing constant (h).
|
|
18
|
- Collision Avoidance Modeling
- Where kj is jam density (veh/km) and un-1(t) is
speed of vehicle n-1 at time t (km/h). This deceleration level is
assumed to be equal to μfbηbg, where
μ is the coefficient of roadway friction, fb is the
driver brake pedal input [0,1], ηb is the brake
efficiency [0,1], and g is the gravitational acceleration (9.8067 m/s2).
|
|
19
|
- Vehicle Acceleration Modeling
- Vehicle acceleration is governed by vehicle dynamics
- Vehicle dynamics models compute the maximum vehicle acceleration levels
from the resultant forces acting on a vehicle.
- where Fn(t) is vehicle tractive force (N), Rn(t)
is total resistance force (N), mn is vehicle mass (kg),fp is
the driver throttle input [0,1], β is the gear reduction factor
(unitless), ηd is the driveline efficiency (unitless), Pn
is the vehicle power (kW), m’n is the mass of vehicle n on
its tractive axle (kg), g is the gravitational acceleration (9.8067 m/s2),
μ is the coefficient of friction (unitless), ρ is the air
density at sea level (1.2256 kg/m3), Cd is the
vehicle drag coefficient (unitless), Ch is the altitude
correction factor (unitless), Af is the vehicle frontal area
(m2), cr0 is the rolling resistance constant
(unitless), cr1 is the rolling resistance constant (h/km), cr2
is the rolling resistance constant (unitless), and G(t) is the roadway
grade at instant t (unitless).
|
|
20
|
|
|
21
|
|
|
22
|
- Step 1: If the spacing between the subject and lead vehicle is beyond
the car-following threshold proceed to step 3, otherwise proceed with
step 2.
- Step 2: Estimate the vehicle at instant t+∆t using the RPA
car-following model and proceed to step 4.
- Step 3: Using DP, the optimum vehicle speed trajectory over the
look-ahead distance (do) is estimated considering a spatial
discretization of length ds (stage length).
- Step 4: Move the vehicle and then go back to step 1 at the conclusion of
the time step Δt; otherwise end the simulation at t=T.
|
|
23
|
|
|
24
|
- Car-following spacing threshold
- Car-following model parameters
- Free-flow speed, Jam density, Speed-at-capacity, and capacity
- Vehicle data
- Powertrain related data, fuel economy data
- Roadway topography
- Real-time GPS data
- Lead vehicle location data (or spacing data)
|
|
25
|
- Three Test Vehicles
- 2011 Toyota Camry (22/33 mpg)
- 2008 Chevy Tahoe (14/20 mpg), and
- 2008 Chevy Malibu Hybrid (24/32 mpg)
- Tested Two Lead Vehicle Trajectories (14miles)
- I-81 SB Field Data (2007 Malibu Manual Driving)
- I-81 SB Eco-Driving Speed Profile (2011 Camry)
- Tested different car-following parameters
- Car-following threshold: 100m, 50m, and 30m
- Throttle level: 100%, 60%, and 40%
- Fixed vs. dynamic threshold
|
|
26
|
|
|
27
|
|
|
28
|
- 20.5 mpg, average spacing=196m,
maximum spacing= 457m
|
|
29
|
- If Spacing > max. spacing (100m) then use car-following model
- 17.2 mpg, average spacing = 48m,
maximum spacing = 133m
|
|
30
|
|
|
31
|
- Car-following only - 23.7 mpg
- ECC mode – 24.6 mpg
|
|
32
|
- Study shows that the proposed system can save fuel significantly
consumption maintaining reasonable vehicle spacing
- Toyota Camry: 27% fuel saving and average spacing: 48m along I-81
- Vehicle operations at lower power demands significantly enhance vehicle
fuel economy (up to 49%)
- Not as significant as the use of the ECC (improved fuel economy up to
82%).
- ECC equipped vehicles benefit following vehicles
- Following vehicles will benefit by just following the lead vehicle.
- There is a need to quantify the potential benefits of using the proposed
system at a network-wide level.
|
|
33
|
|
Notes
