GPU-Accelerated Subsystem-Based ADMM for Large-Scale Interactive Simulation

In this paper, we implement the GPU-accelerated subsystem-based Alternating Direction Method of Multipliers (SubADMM) for interactive simulation. The challenging objective for interactive simulations is to deliver realistic results under tight performance, even for large-scale scenarios. We aim to achieve this by exploiting the parallelizable nature of SubADMM to the fullest extent. We introduce a new subsystem division strategy to make SubADMM ‘GPU friendly' along with custom kernel designs and optimization regarding efficient memory access patterns. We successfully implement the GPUaccelerated SubADMM and show the accuracy and speed of the framework for large-scale scenarios, highlighted with an interactive ‘Hand demo’ scenario. We also show improved robustness and accuracy compared to other state-of-the-art interactive simulators with several challenging scenarios that introduce large-scale ill-conditioned dynamics problems.

Authors

• Harim Ji
• Hyunsu Kim
• Jeongmin Lee 0002
• Somang Lee
• Seoki An
• Jinuk Heo
• Youngseon Lee
• Yongseok Lee
• Dongjun Lee

Keywords

• Hands
• Accuracy
• Scalability
• Memory management
• Graphics processing units
• Programming
• Robustness
• Robotics and automation
• Kernel
• Optimization
• Interactive
• Parallelization
• Dynamic Problem
• Challenging Scenarios
• Large-scale Scenarios
• Kernel Images
• Ill-conditioned Problem
• Detailed Results
• Contact Point
• Lagrange Multiplier
• Rigid Body
• Friction Coefficient
• Iteration Step
• Mass Matrix
• Number Of Bodies
• Slack Variables
• Hard Constraints
• Physical Simulation
• Soft Constraints
• Coriolis Force
• Penalty Weight
• Number Of Subsystems
• Inverse Reinforcement Learning
• Vertical Stacking
• Deformable Body
• Joint Type
• Complementarity Problem
• Scalable
• Class Constraints
• Supplementary Video