%0 Conference Paper %F Oral %T Staircase Traversal via Reinforcement Learning for Active Reconfiguration of Assistive Robots %+ Lab-STICC_IMTA_CID_IHSEV %+ Département Informatique (IMT Atlantique - INFO) %+ IMT Atlantique (IMT Atlantique) %+ Laboratoire des sciences et techniques de l'information, de la communication et de la connaissance (Lab-STICC) %+ Unité d'Informatique et d'Ingénierie des Systèmes (U2IS) %+ Flowing Epigenetic Robots and Systems (Flowers) %A Mitriakov, Andrei %A Papadakis, Panagiotis %A Nguyen, Sao Mai %A Garlatti, Serge %Z The present work is performed in the context of the leadership program M@D (Chaire Maintien@Domicile), project REACT (Physical Interaction Models of Companion Robots) financed by Brest Mtropole and the region of Brittany (France) and supported by project VITAAL (Vaincre l’Isolement par les TIC pour l’Ambient Assisted Living), cofinanced by European Regional Fund (FEDER). %< avec comité de lecture %B IEEE World Congress on Computational Intelligence %C Glasgow, United Kingdom %8 2020-07-19 %D 2020 %K Neural networks %K Learning-based control %K Cognitive robotics %K Obstacle negotiation %K Active stability %K Reinforcement learning %Z Computer Science [cs]/Artificial Intelligence [cs.AI] %Z Computer Science [cs]/Machine Learning [cs.LG] %Z Computer Science [cs]/Human-Computer Interaction [cs.HC] %Z Statistics [stat]/Machine Learning [stat.ML]Conference papers %X Assistive robots introduce a new paradigm for developing advanced personalized services. At the same time, the variability and stochasticity of environments, hardware and unknown parameters of the interaction complicates their modelling , as in the case of staircase traversal. For this task, we propose to treat the problem of robot configuration control within a reinforcement learning framework, using policy gradient optimization. In particular, we examine the use of safety or traction measures as a means for endowing the learned policy with desired properties. Using the proposed framework, we present extensive qualitative and quantitative results where a simulated robot learns to negotiate staircases of variable size, while being subjected to different levels of sensing noise. %G English %2 https://imt-atlantique.hal.science/hal-02676585v1/document %2 https://imt-atlantique.hal.science/hal-02676585v1/file/main.pdf %L hal-02676585 %U https://imt-atlantique.hal.science/hal-02676585