Eco-Drive Experiment on Rolling Terrain for Fuel Consumption Optimization
Chapter 6. Conclusions and Future Research
This study proposes an eco-drive approach that consists of two speed controllers used in our experiment. The first is an upper-level controller that is responsible for trajectory planning to generate optimal speed profiles. The algorithm using the RPMP is computationally efficient and applicable in real time. The secondary controller adjusts speed commands in real time for enhanced performance and contains a typical PID controller for vehicle speed following.
This study further tests these controllers and algorithms in real-world scenarios using an innovative CAV platform to better understand the algorithm performance. The proposed eco-drive system is compared against conventional constant speed cruise control on a total of seven road segments over 47 miles. The number of repetitions for each test is confirmed by statistical tests. The results confirm the fuel saving benefits of eco-drive. Experimental data show that more than 20 percent of fuel consumption can be saved on certain terrains. This research also breaks down test segments into shorter subsegments for further statistical analysis. Detailed investigation reveals the following major findings:
- The benefit of the proposed eco-drive system ranges from 3.3 percent to 21.2 percent. The magnitude of benefit is affected by hill length and slope grade.
- Conventional constant speed cruise control is fuel inefficient. It tends to apply gas on the uphill segment and then brake on the downhill segment. Hence, the replacement of conventional cruise control with the proposed eco-drive system on rolling terrains is very beneficial.
- Eco-drive reduces the amount of acceleration, leading to less braking. Less braking and smaller brake percentages translate to a more effective transformation between potential and kinetic energy, thus less additional fuel is needed to provide the extra kinetic energy.
- Detailed analysis through linear models also reveals the main contributors to the fuel savings potentially obtained from eco-drive. This conclusion can enable a rough estimate of the fuel saving potential of given roadways and help State DOTs to identify roadways where V2I-enabled eco-drive should be implemented.
The linear model can also be used to roughly estimate the benefits of eco-drive for any roadway segments. This model can enable rough estimation of the fuel saving potential of given roadways and help State DOTs to identify locations where eco-drive should be implemented. The algorithm and the experiment can also support OEMs in developing and marketing this technology to reduce fuel consumption and emissions in the future.
In terms of suggested future studies, there are many directions in which to build from this study.
- Field data on more experimental sites can be collected to examine eco-drive performance on a wide range of terrain conditions, and then more detailed subsegment analysis can be conducted.
- Advanced algorithms may also consider the existence of a front vehicle (via vehicle-to-vehicle information) and downstream traffic congestion through addition of a speed harmonization component to the algorithm (Ma et al., 2016).
- The proposed algorithm used speed commands to control the vehicle and did not exert control over the vehicle transmission. The experiment may be repeated with algorithm output that controls the transmission in addition to speed or throttle and an experimental vehicle with the capability of accepting commands for transmission state.
- The current optimization method for synthesizing the optimal vehicle speed doesn't consider the gear shift schedule. This will cause some inaccuracy in estimating the fuel efficiency improvement since engines can only work at the speed dictated by the transmission gear shift schedule. One suggestion is to include the nonlinear gear shift schedule in the optimization method.
- To further improve the energy efficiency, co-optimization of the vehicle speed and the transmission gear shift schedule could be conducted simultaneously. This is different than the previous approach, where only the existing gear shift schedule was included in the optimization method.
- The proposed algorithm relies on an approximate vehicle dynamics model, and thus actual vehicles may not follow the commanded speed profiles well. Incorporating machine learning and artificial intelligence components in future algorithms will allow vehicles to plan the most suitable trajectories for themselves by using data from earlier operations.
- This study focused on the benefits to a single vehicle running the experimental eco-drive algorithm. Further research is still needed to study the impact that the subject vehicle has on the traffic following it, including automated and nonautomated vehicles.
