Projects / Programmes
Multiobjective discovery of driving strategies for autonomous vehicles
| Code |
Science |
Field |
Subfield |
| 2.07.07 |
Engineering sciences and technologies |
Computer science and informatics |
Intelligent systems - software |
| Code |
Science |
Field |
| P176 |
Natural sciences and mathematics |
Artificial intelligence |
| Code |
Science |
Field |
| 1.02 |
Natural Sciences |
Computer and information sciences |
real-time optimization; driving strategies; traveling time; fuel consumption; multiobjective optimization; black-box approach
Researchers (1)
| no. |
Code |
Name and surname |
Research area |
Role |
Period |
No. of publicationsNo. of publications |
| 1. |
31049 |
PhD Erik Dovgan |
Computer science and informatics |
Head |
2016 - 2018 |
143 |
Organisations (1)
Abstract
There have been several investments in the autonomous vehicle driving technologies by automotive companies recently. The focus of these technologies is mostly on determining the vehicle surroundings in order to increase the driving safety and avoid collisions. In addition, several researches have developed optimization algorithms for discovering driving strategies, which optimize also other driving aspects such as the traveling time, fuel consumption, emissions etc.
The proposed project aims at developing an algorithm for discovering driving strategies that, on the one hand, could be deployed in real vehicles, and, on the other hand, would be also suitable for solving hard optimization problems such as real-time discovery of driving strategies with short traveling time and low fuel consumption.
The leader of this project has already developed an algorithm for discovering driving strategies (MODS). However, this algorithm does not take into account real-time requirements (the algorithm has to be speed up) and interaction with other vehicles (appropriate procedures for such an interaction have to be added). In addition, it has been only tested on a computer-based simulator that does not include any driving-dedicated hardware components (e.g., vibration simulator) or data about other vehicles on the route. Moreover, the algorithm does not take human aspects into account when searching for driving strategies. Consequently, even if a good driving strategy is found, it could not be suitable (e.g., comfortable enough) for the users.
During the proposed project, the MODS algorithm will be significantly enhanced and properly evaluated in order to be ready for real-life deployment and usage. More precisely, several possibilities for increasing its efficiency will be studied and the most suitable will be implemented in order to obtain an algorithm that finds driving strategies in real time. In addition, human driving characteristics will be obtained, analyzed and properly included in MODS to obtain human-like driving strategies. Moreover, the information about other vehicles on the route (e.g., the position and velocity of the neighbor vehicles) will be integrated in MODS. Furthermore, MODS will be integrated and tested in a near-real-life environment, consisting of a car driving simulator provided by the company that will co-fund the proposed project. This simulator is used for training, education and evaluation of the drivers. It consists of a dedicated hardware platform and a software simulator, and is able to simulate traffic involvement, safety aspects, unexpected situations, traffic fluidity, fuel consumption, etc.
Such an enhanced real-time optimization algorithm, integrated in the car driving simulator will bring several benefits for the co-funding company. On the one hand, this integration will enable to compare the discovered driving strategies with the vehicle driving of the users of the simulator. More precisely, the discovered driving strategies will represent the optimal solution to which the users should aim at. On the other hand, the co-funding company aims at developing a solution for autonomous vehicle driving based on the enhanced MODS algorithm, which could be sold to automotive companies and thus deployed in real vehicles.
Significance for science
The area of autonomous driving is very active and rapidly advancing. In addition to the perception of the environment, the autonomous driving also focuses on the appropriate selection of the best driving strategy. A new approach for discovering driving strategies has been developed within this project, which combines the optimization of driving objectives with the characteristics of human driving. These two driving aspects are typically taken into account independently, since the existing approaches focus either on optimizing the traveling time, fuel consumption and/or other objectives, or on discovering driving strategies similar to human driving strategies. The developed approach is very general, as it enables the definition of additional features that must be included in the driving strategies and additional objectives that the driving strategies need to meet. Moreover, the generality of the developed approach makes it possible to apply or upgrade it in other research areas. The results of the project are a step towards the appropriate management of heterogeneous requirements for autonomous driving in the form of preferred driving characteristics and objectives, which will enable a comprehensive approach for the optimization of driving strategies.
Significance for the country
The driving strategies found with the multiobjective optimization will improve driver training and assessment, which will increase the added value of the driving simulator. In addition, the developed optimization algorithm represents the core of a new solution for the autonomous driving of vehicles, whose development is carried out by the co-funding company. The target market for this solution consists of the automotive and other companies operating in the field of autonomous driving. Successful penetration of the co-funding company in this field already enables it to expand both in terms of penetration into new markets and in terms of additional employment. In addition, the project has established a long-term cooperation between researchers and development personnel in the field of the autonomous driving, which will in the future enable and accelerate the development of new market-relevant solutions for autonomous driving.
Most important scientific results
Interim report,
final report
Most important socioeconomically and culturally relevant results
Interim report,
final report
