Martin Riedmiller

About

Martin Riedmiller is a pioneering researcher in reinforcement learning (RL) and robotics, whose work has fundamentally shaped how autonomous systems learn complex behaviors from experience. Based at DeepMind, his research spans deep reinforcement learning, robot learning, and sparse reward problems — areas where his contributions have earned thousands of citations and widespread recognition across the AI community. Riedmiller is perhaps best known for bridging the gap between theoretical RL and real-world robotics. His landmark 2017 paper on leveraging demonstrations for deep RL with sparse rewards (510 citations) demonstrated how combining human demonstrations with autonomous exploration enables robots to master challenging manipulation tasks. His development of Scheduled Auxiliary Control (SAC-X) further addressed the notoriously difficult sparse reward problem, enabling agents to learn complex behaviors entirely from scratch. His work on skill embedding spaces (190 citations) advanced transfer learning in robotics, allowing agents to generalize across related tasks efficiently. An early advocate of learning-based robot soccer — documented through his Karlsruhe Brainstormers projects — Riedmiller has consistently pushed RL toward practical deployment, including teaching a real car to drive in just 20 minutes. His body of work collectively represents a sustained, impactful effort to make intelligent autonomous systems a tangible reality.

Research Focus

Key Achievements

Top Papers

1. 1
   Leveraging Demonstrations for Deep Reinforcement Learning on Robotics Problems with Sparse Rewards

   510 citations · 2017
2. 2
   Reinforcement learning for robot soccer

   266 citations · 2009
3. 3
   Learning an Embedding Space for Transferable Robot Skills

   190 citations · 2018
4. 4
   Learning by Playing - Solving Sparse Reward Tasks from Scratch

   155 citations · 2018
5. 5
   Multimodal deep learning for robust RGB-D object recognition

   133 citations · 2015
6. 6
   Data-efficient Deep Reinforcement Learning for Dexterous Manipulation

   118 citations · 2017
7. 7
   Calculating the Perfect Match: An Efficient and Accurate Approach for Robot Self-localization

   101 citations · 2006
8. 8
   Karlsruhe Brainstormers - A Reinforcement Learning approach to robotic soccer

   100 citations · 2001
9. 9
   Learning to Drive a Real Car in 20 Minutes

   92 citations · 2007
10. 10
    Acquiring visual servoing reaching and grasping skills using neural reinforcement learning

    64 citations · 2013

Key Collaborators

Contact & Links