Doctoral Researcher: Continual Reinforcement Learning

Aalto; Espoo; On-site; EUR 32,000 - 38,000; Full time

Benefits offered by this job

Fully funded position

Fixed 4-year term

Competitive salary

International collaborations

Job summary

Aalto University invites applications for two Doctoral Researchers in Reinforcement Learning within the Cyber-physical Systems Group of the School of Electrical Engineering. The ERC Starting Grant project aims to address continual RL for real-world systems that do not reset and keep changing, offering a four-year, fully funded research environment.

Position 1 emphasizes trajectory-centric optimization, Position 2 focuses on adaptive policies, with joint integration into a single RL algorithm.

Qualifications

• Master’s degree in computer science, mathematics, electrical engineering, or related fields.
• Fluent written and verbal communication skills in English.
• Ability to work both independently and collaboratively as part of a research team.

Responsibilities

• Develop theory for trajectory-centric stochastic optimization under non-ergodic dynamics.
• Implement practical RL algorithms that optimize long-term performance of agents.
• Develop change-detection algorithms to infer when a policy must be adapted.
• Implement policy-adaptation algorithms with safety guarantees.
• Integrate advances into state-of-the-art RL algorithms.
• Validate algorithms in simulation and hardware environments.

Skills

Reinforcement learning

Trajectory-centric optimization

Adaptive policies

Non-ergodic dynamics

Education

Master’s degree in CS/Math/EE

Job description

Aalto University invites applications for two Doctoral Researchers in Reinforcement Learning within the Cyber-physical Systems Group of the School of Electrical Engineering. The ERC Starting Grant project aims to address continual RL for real-world systems that do not reset and keep changing, offering a four-year, fully funded research environment.

Position 1 emphasizes trajectory-centric optimization, Position 2 focuses on adaptive policies, with joint integration into a single RL algorithm.