Robust constraint-consistent learning

Many everyday human skills can be framed in terms of performing some task subject to constraints imposed by the environment. Constraints are usually unobservable and frequently change between contexts. In this paper, we present a novel approach for learning (unconstrained) control policies from movement data, where observations are recorded under different constraint settings. Our approach seamlessly integrates unconstrained and constrained observations by performing hybrid optimisation of two risk functionals. The first is a novel risk functional that makes a meaningful comparison between the estimated policy and constrained observations. The second is the standard risk, used to reduce the expected error under impoverished sets of constraints. We demonstrate our approach on systems of varying complexity, and illustrate its utility for transfer learning of a car washing task from human motion capture data.

Authors

• Matthew J. Howard 0001
• Stefan Klanke
• Michael Gienger
• Christian Goerick
• Sethu Vijayakumar

Keywords

• Robustness
• Humans
• Page description languages
• Motion control
• Intelligent robots
• USA Councils
• Constraint optimization
• Anthropomorphism
• Mouth
• Control systems
• Standard Risk
• Linear Model
• Model Parameters
• Horizontal Axis
• Receptive Field
• Direct Approach
• Radial Basis Function
• Hybrid Approach
• Observable Variables
• Environmental Constraints
• Policy Variables
• Effect Of Constraints
• Policy Learning
• Policy Model
• Hard Constraints
• Data Constraints
• Look For Solutions
• Component Of Policy
• Constraint Matrix
• Direct Regression
• Dimensional Constraints
• Single Constraint
• Limit Cycle
• Toy Example
• State Space
• Least-squares Fitting
• Eigenvalues
• Form Of Constraints
• Projection Matrix
• Optimal Weight