Abstract:Despite recent successes of reinforcement learning (RL), it remains a challenge for agents to transfer learned skills to related environments. To facilitate research addressing this problem, we propose CausalWorld, a benchmark for causal structure and transfer learning in a robotic manipulation environment. The environment is a simulation of an open-source robotic platform, hence offering the possibility of sim-to-real transfer. Tasks consist of constructing 3D shapes from a given set of blocks - inspired by how children learn to build complex structures. The key strength of CausalWorld is that it provides a combinatorial family of such tasks with common causal structure and underlying factors (including, e.g., robot and object masses, colors, sizes). The user (or the agent) may intervene on all causal variables, which allows for fine-grained control over how similar different tasks (or task distributions) are. One can thus easily define training and evaluation distributions of a desired difficulty level, targeting a specific form of generalization (e.g., only changes in appearance or object mass). Further, this common parametrization facilitates defining curricula by interpolating between an initial and a target task. While users may define their own task distributions, we present eight meaningful distributions as concrete benchmarks, ranging from simple to very challenging, all of which require long-horizon planning as well as precise low-level motor control. Finally, we provide baseline results for a subset of these tasks on distinct training curricula and corresponding evaluation protocols, verifying the feasibility of the tasks in this benchmark.

ROS-Causal: A ROS-based Causal Analysis Framework for Human-Robot Interaction Applications

Experimental Evaluation of ROS-Causal in Real-World Human-Robot Spatial Interaction Scenarios

Causal Discovery of Dynamic Models for Predicting Human Spatial Interactions

Efficient Causal Discovery for Robotics Applications

An Introduction to Causal Inference Methods for Observational Human-Robot Interaction Research

From Continual Learning to Causal Discovery in Robotics

A Human-Centered Approach for Bootstrapping Causal Graph Creation

Evaluating Temporal Observation-Based Causal Discovery Techniques Applied to Road Driver Behaviour

Learning Low-Level Causal Relations using a Simulated Robotic Arm

Towards a Causal Probabilistic Framework for Prediction, Action-Selection & Explanations for Robot Block-Stacking Tasks

Enhancing Causal Discovery from Robot Sensor Data in Dynamic Scenarios

CAnDOIT: Causal Discovery with Observational and Interventional Data from Time-Series

Towards Probabilistic Causal Discovery, Inference & Explanations for Autonomous Drones in Mine Surveying Tasks

Message Flow Analysis with Complex Causal Links for Distributed ROS 2 Systems

Simulation-Based Counterfactual Causal Discovery on Real World Driver Behaviour

Causal Reinforcement Learning for Optimisation of Robot Dynamics in Unknown Environments

ccaROS: A ROS Node for Cognitive Collaborative Architecture for an Intelligent Wheelchair

CausalWorld: A Robotic Manipulation Benchmark for Causal Structure and Transfer Learning

Causal Counterfactuals for Improving the Robustness of Reinforcement Learning

Extending Structural Causal Models for Autonomous Embodied Systems

CASPR-ROS: A Generalised Cable Robot Software in ROS for Hardware