Reinforcement Learning in Robotics

Intrinsic Goals for Autonomous Agents: Hierarchical Exploration Captures Future Volitional Action in Drosophila

This study uses intrinsic motivation algorithms to create agents that mimic Drosophila exploration behavior and neural representations of future actions.

Inferring the evolving objectives that organize animal behavior

This paper develops inverse reinforcement learning to infer evolving objectives from animal behavior and shows it can identify time-varying objectives in mice.

Manifold Agentic Reasoning: Extending Agentic POMDPs and Post-Training Reasoning to Riemannian State and Reasoning Spaces

This paper introduces a geometric framework for agentic reasoning on curved manifolds, showing improved performance in a tissue repair task compared to flat-space methods.

Interpreting Rewards from Inverse Reinforcement Learning

This paper introduces a framework for interpreting reward functions from inverse reinforcement learning, demonstrating its application to multi-agent social interactions.

Efficient Learning of Predictive Maps for Flexible Planning

This paper introduces SR-IS, a model that learns policy-independent predictive maps for flexible planning, outperforming existing models in planning tasks and explaining human replanning biases.

Task-space dimensions guide human exploration in complex environments

The paper shows humans use a structured exploration strategy in complex environments by systematically sampling task dimensions, improving learning efficiency.

Default Feature Representations of the Cognitive Map

This paper introduces Default Feature Representations (DFR) for updating predictive cognitive maps, showing it outperforms baselines in replanning tasks and captures grid cell remapping.

Adaptive integration of model-based and model-free strategies in human reinforcement learning of reachable space

This study shows humans adaptively use model-based and model-free strategies in learning to reach around obstacles in reachable space, with increased model-free reliance as learning progresses.

Why motor learning involves multiple systems: an algorithmic perspective

This paper proposes a computational framework where the cortico-cerebellar network learns low-dimensional action embeddings and the basal ganglia use reinforcement learning to learn value and policies, demonstrating its relevance to motor learning.

A model-centered introduction to curiosity

The paper compares curiosity models using experimental results and mathematical formulations to understand how motivation and exploration evolve in uncertain environments.

Individual and social contributions to learning of mice in Intellicages: the general framework and mean field theory

This paper introduces a framework combining point processes with reinforcement learning to analyze individual and group learning in mice, finding that group effects are twice as strong as individual learning.

Inferring learning rules during de novo task learning

This study infers reinforcement learning rules from single-animal behavior during de novo task learning in mice, finding that policy-gradient-like rules outperform classical temporal-difference algorithms.

Human Strategy Adaptation in Reinforcement Learning Resembles Policy Gradient Ascent

This paper introduces DynamicRL, a method to estimate how humans adapt their reinforcement learning strategies over time, showing parameter updates resemble policy gradient ascent.

Horizontally Tiled Network of Cortico-Basal Ganglia Modules Performs Reinforcement Learning

This study demonstrates that reinforcement learning can be achieved through horizontal signal transmission between horizontally tiled cortico-basal ganglia modules, leveraging differences in basal ganglia pathway projection ranges.

Advances in deep reinforcement learning enable better predictions of human behavior in time-continuous tasks

This study uses advanced deep reinforcement learning models to predict human motor responses in time-continuous tasks, finding that more complex models achieve higher prediction accuracy.

Multi-model reinforcement learning with online retrospective change point detection

This paper introduces a reinforcement learning agent that adapts to changing environments by detecting change points retrospectively and managing multiple local models.

On the duality of pain and pleasure processing: Why two dimensions of valence may be better than one

This paper shows that separate pain and pleasure processing systems outperform unified reward systems in dynamic environments through modular reinforcement learning architectures.

Modeling Complex Animal Behavior with Latent State Inverse Reinforcement Learning

This paper introduces dDIRL to model complex animal behavior over time, identifying internal states and reward functions from observed navigation in water-starved mice.

Unveiling value functions in social cognition with multi-agentinverse reinforcement learning

This paper introduces MAIRL, a multi-agent inverse reinforcement learning framework that decomposes joint value functions into individual value maps and interaction terms, revealing interpretable social value representations in mice and primates.

Capturing Optimal and Suboptimal behavior Of Agents Via Structure Learning Of Their Internal Model

This paper introduces a framework to infer agents' internal models from behavior, revealing that suboptimal decisions arise from interactions between learning rules and environment representations in rats.

Extinction burst could be explained by curiosity-driven reinforcement learning

This study shows that extinction bursts in mice can be explained by curiosity-driven reinforcement learning, linking intrinsic motivation to behavioral responses.

Online reinforcement learning of state representation in recurrent network: the power of random feedback and biological constraints

This study shows that RNNs can learn task structure with random feedback and biological constraints without needing offline updates or eligibility traces.

Inverse Reinforcement Learning to Study Motivation in Mouse Behavioral Paradigms

This paper applies inverse reinforcement learning to directly infer the underlying motivations driving mouse behavior in behavioral paradigms.

Synthesising integrated robot behaviour through reinforcement learning for homeostasis

This paper introduces an algorithm inspired by neuroscience to enable robots to exhibit integrated survival behaviors through reinforcement learning focused on homeostasis.

Neuron-level prediction and noise can implement flexible reward-seeking behavior

This paper demonstrates that neural networks can exhibit reward-seeking behavior through local predictive updates and internal noise, enabling autonomous and adaptable interaction with environments.

A meta reinforcement learning account of behavioral adaptation to volatility in recurrent neural networks

The paper shows that recurrent neural networks trained with medium or meta-volatility regimes adapt better to environmental volatility than those trained in low volatility.

Generative Adversarial Implicit Successor Representation

This paper introduces a method to implicitly encode Successor Representations using a GAN, enabling action decisions without explicit probability calculations.

Combining Backpropagation with Equilibrium Propagation to improve an Actor-Critic RL framework

This paper combines equilibrium propagation with backpropagation in an actor-critic reinforcement learning framework, showing improved consistency in high-reward behavior on control tasks.

Reward Bases: Instantaneous reward revaluation with temporal difference learning

The paper introduces an extension to temporal difference learning that enables instantaneous adaptation to changing reward functions without retraining.

Validating the Representational Space of DeepReinforcement Learning Models of Behavior withNeural Data

This paper validates the neural representations of Deep RL agents against animal neural data, showing similar geometry in latent spaces and the role of temporal information.