Imitation Learning with Inconsistent Demonstrations through Uncertainty-based Data Manipulation

Aleatoric uncertainty estimation, based on the observed training data, is applied for the detection of conflicts in a demonstration data set. The particular focus of this paper is the resolution of conflicting data resulting from scenarios with equivalent action choices, such as obstacle avoidance, path planning or multiple joint configurations. In terms of the estimated uncertainty, the proposed algorithm aims to decrease this otherwise irreducible value through direct alteration of the accrued data set and to provide data that a policy-learning neural network is able to fit appropriately. The proposed algorithm was validated with real robot scenarios while learning from inconsistent demonstrations, where the resulting policies consistently achieved their prescribed objectives. A video showing our method and experiments can be found at: https://youtu.be/oGYnzlW9Ncw.

Authors

• Peter Valletta
• Rodrigo Pérez-Dattari
• Jens Kober

Keywords

• Uncertainty
• Conferences
• Neural networks
• Clustering algorithms
• Estimation
• Training data
• Inference algorithms
• Imitation Learning
• Neural Network
• Focus Of This Paper
• Uncertainty Estimation
• Path Planning
• Obstacle Avoidance
• Aleatoric Uncertainty
• Covariance Matrix
• State Space
• Conflict Resolution
• Continuous Action
• Feed-forward Network
• Mean Prediction
• Gaussian Mixture Model
• Actual Dimensions
• Epistemic Uncertainty
• Task Space
• Uncertainty Region
• DBSCAN Algorithm