Feedback motion planning for liquid pouring using supervised learning

We present a novel motion planning algorithm for pouring a liquid body from a source to a target container. Our approach uses a receding-horizon optimization strategy that considers liquid dynamics and various other constraints. To handle liquid dynamics without costly fluid simulations, we use a neural network to infer a set of key liquid-related parameters from the observation of the current liquid configuration. To train the neural network, we generate a dataset of successful pouring examples using stochastic optimization in a problem-specific search space. These parameters are then used in the objective function for trajectory optimization. Our feedback motion planner achieves real-time performance, and we observe a high success rate in our simulated 2D and 3D liquid pouring benchmarks.

Authors

• Zherong Pan
• Dinesh Manocha

Keywords

• Liquids
• Trajectory
• Containers
• Planning
• Heuristic algorithms
• Dynamics
• Optimization
• Supervised Learning
• Path Planning
• Motion Feedback
• Neural Network
• Objective Function
• Search Space
• Stochastic Optimization
• Simulated Fluid
• Trajectory Optimization
• Planning Algorithm
• Dynamic Liquid
• Training Dataset
• Flow Velocity
• Laminar Flow
• Rigid Body
• Turbulent Flow
• Free Surface
• Navier Stokes Equations
• Reward Function
• Spline Interpolation
• Imitation Learning
• Number Of Trajectories
• Quadratic Curve
• Humanoid Robot
• Successional Trajectories
• Liquid Simulations
• Dynamic Obstacles
• Granular Material
• Low-dimensional Feature
• Real-life Systems