Passivity-Based Haptic Rendering and Interactive Simulation Framework with Multi-Point Contacts

Abstract

In this thesis, we propose a haptic rendering and interactive simulation framework based on passive midpoint integrator (PMI) which can enforce discrete-time passivity of the simulation effectively in practice while retaining real-time interactivity from its being non-iterative. We derive this PMI simulation for mechanical systems both in generalized coordinates (i.e., in $\Re^n$) and also in maximal coordinates (i.e., in SE(3)), with some potential actions as well to implement joint articulation, constraints, compliance, etc. We also formulate both constraint-based contact solving and penalty-based contact solving for the PMI simulation, to resolve multi-point contact stably and interactively. To address/improve numerical stability and accuracy when too many contacts are engaged, we also propose a data-driven learning-based contact clustering, which consists of the multilayer perceptron (MLP) network and the constraint-based optimization contact solver.

We manifest the advantage of the proposed PMI-based simulation framework with some illustrative examples: 1) haptic rendering of peg-in-hole task, where very light/stiff articulated objects can be simulated with multi-point contact

2) haptic interaction with flexible beam, where marginally-stable/lossless behavior (i.e., vibration) can be stably emulated

3) under-actuated tendon-driven hand grasping, where mixed maximal-generalized coordinates are used with very light/stiff fingers

and 4) multiuser haptic interaction under delayed communication network, where peer-to-peer control architecture and passive consensus control are employed. We also verify the performance of the proposed data-driven/learning-based contact clustering framework, against the experimental contact data, compare it with other techniques/simulators, and show that there is significant (or meaningful) enhancement in the accuracy.

본 논문에서는 수동성 기반의 다중접촉 햅틱렌더링 및 실시간 시뮬레이션 기법을 제안한다. 제안된 기법은 1) 일반화좌표계의 비선형 라그랑지 동역학의 이산시간 수동성을 보장하는 Passive Midpoint Integration, 2) 최대좌표계의 SE(3) 강체들과 그들 사이의 구속력을 수동성을 근사적으로 보장하도록 하는 Passive Midpoint Integration, 3) 물체들 사이의 다중접촉을 풀기 위한 컨스트레인트 기반 및 페널티기반의 다중접촉 시뮬레이션 기법, 4) 가상커플링을 이용한 햅틱렌더링기법, 및 마지막으로 5) 다중접촉의 정확도 향상을 위한 실험데이터 및 기계학습기반의 접촉 클러스터링 기법으로 구성된다.

제안된 기법의 유효성을 검증하기 위하여 진동빔 시뮬레이션, 부족구동 텐던로봇 손, 다중접촉 펙인홀 작업 및 통신시간지연하의 원격사용자간의 협동 펙인홀 햅틱 상호작용에 대한 실험 결과를 제시하며, 다중접촉 시뮬레이션의 성능비교평가 결과를 제시한다.