Master's Thesis - inverse reinforcement learning for autonomous driving
The document proposes a master's thesis on applying inverse reinforcement learning to autonomous driving. Inverse reinforcement learning infers reward functions from expert demonstrations, allowing reinforcement learning to be applied to problems where the reward function is unknown. The thesis would deepen the theory of inverse reinforcement learning and test algorithms in simulated vehicle environments. Over 5-6 months, the student would acquire expertise in reinforcement learning, proficiency in applying inverse reinforcement learning to autonomous driving, and experience applying machine learning to complex systems. The student should have programming skills and knowledge of machine learning, linear algebra, and vehicle control systems.
![MASTERTHESISPROPOSAL
Inverse Reinforcement Learning for Autonomous Driving
MASTERTHESISPROPOSAL
Reinforcement Learning [1] (RL) is an emerging ?eld of Arti?cial Intelligence (AI) that is giving
extraordinary results in different applications.
One of such applications is Autonomous Driving, but to apply RL to this task an accurate choice of the
reward function is needed.To overcome this issue, one solution is to infer the reward function applying
Machine Learning (ML) techniques to some examples provided by experts. For example, a driver can
show how to do a speci?c maneuver and a ML algorithm extract the objective function maximized by
the driver behaviour.This method is known as Inverse Reinforcement Learning [2] (IRL).
The thesis will deepen the theory behind inverse reinforcement learning to analyze the possible
applications of this approach to autonomous driving [3] in a simulated environment [4, 5, 6].
Planned Activities
1. Acquire ?strong theoretical basis on Reinforcement Learning and in particular on Inverse Reinforcement Learning; ?
2. Test the latest algorithms to ＾toy￣ environments;?
3. Adapt and test the most promising algorithms to a complex vehicle model.
Competencies to be acquired
The candidate will acquire:
? Experience with the application of Machine Learning to complex systems.?
? Expertise on the most recent Reinforcement Learning algorithms;?
? Pro?ciency in the application of Inverse Reinforcement Learning algorithms to Autonomous Driving.?
Duration of this Project: 5-6 months.
Check these Links before moving on
[1] Reinforcement Learning: An Introduction
http://incompleteideas.net/book/the-book.html
[2] Inverse Reinforcement Learning
http://ai.stanford.edu/%7Eang/papers/icml00-irl.pdf
[3] Learning Driving Styles for Autonomous Vehicles from Demonstration
http://ais.informatik.uni-freiburg.de/publications/papers/kuderer15icra.pdf
[4] Simulated Car Racing Championship Competition - Software Manual
https://arxiv.org/pdf/1304.1672.pdf
[5] CARLA Simulator
https://github.com/carla-simulator/carla
[6] Microsoft AirSim
https://github.com/Microsoft/AirSim
add-for.com
Who we¨re looking forStudents that are about to get their Master Degree in: computer engineering, mechatronic engineering,
aerospace engineering, automotive engineering, electronic engineering.
Required Skills:
Pro?ciency in at least one programming language (Python, Lua, Matlab/Simulink, C++, Java);
Basic knowledge of machine learning;
Good knowledge of linear algebra;
Basic knowledge of vehicle control systems and vehicle dynamics.
How to contact us
Directly by email to: sonia.cannavo@add-for.com
By LinkedIn: linkedin.com/in/cannav┛-sonia-66a95467](https://image.slidesharecdn.com/thesis22-inversereinforcementlearningforautonomousdriving-190404042557/85/Master-s-Thesis-inverse-reinforcement-learning-for-autonomous-driving-1-320.jpg)
Master's Thesis - inverse reinforcement learning for autonomous driving
- 1. MASTERTHESISPROPOSAL Inverse Reinforcement Learning for Autonomous Driving MASTERTHESISPROPOSAL Reinforcement Learning [1] (RL) is an emerging ?eld of Arti?cial Intelligence (AI) that is giving extraordinary results in different applications. One of such applications is Autonomous Driving, but to apply RL to this task an accurate choice of the reward function is needed.To overcome this issue, one solution is to infer the reward function applying Machine Learning (ML) techniques to some examples provided by experts. For example, a driver can show how to do a speci?c maneuver and a ML algorithm extract the objective function maximized by the driver behaviour.This method is known as Inverse Reinforcement Learning [2] (IRL). The thesis will deepen the theory behind inverse reinforcement learning to analyze the possible applications of this approach to autonomous driving [3] in a simulated environment [4, 5, 6]. Planned Activities 1. Acquire ?strong theoretical basis on Reinforcement Learning and in particular on Inverse Reinforcement Learning; ? 2. Test the latest algorithms to ＾toy￣ environments;? 3. Adapt and test the most promising algorithms to a complex vehicle model. Competencies to be acquired The candidate will acquire: ? Experience with the application of Machine Learning to complex systems.? ? Expertise on the most recent Reinforcement Learning algorithms;? ? Pro?ciency in the application of Inverse Reinforcement Learning algorithms to Autonomous Driving.? Duration of this Project: 5-6 months. Check these Links before moving on [1] Reinforcement Learning: An Introduction http://incompleteideas.net/book/the-book.html [2] Inverse Reinforcement Learning http://ai.stanford.edu/%7Eang/papers/icml00-irl.pdf [3] Learning Driving Styles for Autonomous Vehicles from Demonstration http://ais.informatik.uni-freiburg.de/publications/papers/kuderer15icra.pdf [4] Simulated Car Racing Championship Competition - Software Manual https://arxiv.org/pdf/1304.1672.pdf [5] CARLA Simulator https://github.com/carla-simulator/carla [6] Microsoft AirSim https://github.com/Microsoft/AirSim add-for.com Who we¨re looking forStudents that are about to get their Master Degree in: computer engineering, mechatronic engineering, aerospace engineering, automotive engineering, electronic engineering. Required Skills: Pro?ciency in at least one programming language (Python, Lua, Matlab/Simulink, C++, Java); Basic knowledge of machine learning; Good knowledge of linear algebra; Basic knowledge of vehicle control systems and vehicle dynamics. How to contact us Directly by email to: sonia.cannavo@add-for.com By LinkedIn: linkedin.com/in/cannav┛-sonia-66a95467