QuasiSim: Parameterized Quasi-Bodily Simulators for Dexterous Manipulations Switch
Authors: Xueyi Liu, Kangbo Lyu, Jieqiong Zhang, Tao Du, Li Yi
Summary: We discover the dexterous manipulation switch drawback by designing simulators. The duty needs to switch human manipulations to dexterous robotic hand simulations and is inherently troublesome as a consequence of its intricate, highly-constrained, and discontinuous dynamics and the necessity to management a dexterous hand with a DoF to precisely replicate human manipulations. Earlier approaches that optimize in high-fidelity black-box simulators or a modified one with relaxed constraints solely display restricted capabilities or are restricted by inadequate simulation constancy. We introduce parameterized quasi-physical simulators and a physics curriculum to beat these limitations. The important thing concepts are 1) balancing between constancy and optimizability of the simulation by way of a curriculum of parameterized simulators, and a couple of) fixing the issue in every of the simulators from the curriculum, with properties starting from excessive activity optimizability to excessive constancy. We efficiently allow a dexterous hand to trace advanced and numerous manipulations in high-fidelity simulated environments, boosting the success price by 11%+ from the best-performed baseline. The venture web site is obtainable at https://meowuu7.github.io/QuasiSim/.
