Efficient global penetration depth computation for articulated models

Tian, Hao; Zhang, Xinyu; Wang, Changbo; Pan, Jia; Manocha, Dinesh

doi:10.1016/j.cad.2015.07.007

Cited by 4 publications

(4 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These samples can be drawn guided by an acquisition function in Bayesian optimization (Niculescu et al 2006) or from an expert algorithm (De Raedt, Passerini, and Teso 2018). Active learning has been applied to approximate the boundary of the configuration space (Pan, Zhang, and Manocha 2013;Tian et al 2016;Das, Gupta, and Yip 2017), where the feasible domain of collision constraints is parameterized using kernel SVM. However, this method is limited to rigid or articulated deformation and is not applicable to general 3D deformations.…”

Section: Related Workmentioning

confidence: 99%

Active Learning of Neural Collision Handler for Complex 3D Mesh Deformations

Tan,

Pan,

Smith

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

We present a robust learning algorithm to detect and handle collisions in 3D deforming meshes. Our collision detector is represented as a bilevel deep autoencoder with an attention mechanism that identifies colliding mesh sub-parts. We use a numerical optimization algorithm to resolve penetrations guided by the network. Our learned collision handler can resolve collisions for unseen, high-dimensional meshes with thousands of vertices. To obtain stable network performance in such large and unseen spaces, we progressively insert new collision data based on the errors in network inferences. We automatically label these data using an analytical collision detector and progressively fine-tune our detection networks. We evaluate our method for collision handling of complex, 3D meshes coming from several datasets with different shapes and topologies, including datasets corresponding to dressed and undressed human poses, cloth simulations, and human hand poses acquired using multiview capture systems. Our approach outperforms supervised learning methods and achieves 93.8 − 98.1% accuracy compared to the groundtruth by analytic methods. Compared to prior learning methods, our approach results in a 5.16% − 25.50% lower false negative rate in terms of collision checking and a 9.65% − 58.91% higher success rate in collision handling.

show abstract

Section: Related Workmentioning

confidence: 99%

Active Learning of Neural Collision Handler for Complex 3D Mesh Deformations

Tan,

Pan,

Smith

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The method described here for computing an implicit contact manifold for an articulated body is essentially the same as one described in [30].…”

Section: Projecting Onto the Contact Manifoldmentioning

confidence: 99%

The manifold particle filter for state estimation on high-dimensional implicit manifolds

Koval

Klingensmith

Srinivasa

et al. 2017

2017 IEEE International Conference on Robotics and Automation (ICRA)

View full text Add to dashboard Cite

Abstract-We estimate the state a noisy robot arm and underactuated hand using an Implicit Manifold Particle Filter (MPF) informed by touch sensors. As the robot touches the world, its state space collapses to a contact manifold that we represent implicitly using a signed distance field. This allows us to extend the MPF to higher (six or more) dimensional state spaces. Earlier work (which explicitly represents the contact manifold) only shows the MPF in two or three dimensions [1]. Through a series of experiments, we show that the implicit MPF converges faster and is more accurate than a conventional particle filter during periods of persistent contact. We present three methods of sampling the implicit contact manifold, and compare them in experiments.

show abstract

“…This does not relate to the total required transformation of two dynamic bodies, nor is it meaningful if only the “static” body is free to move. Most works, however, define one of the bodies as the dynamic one in advance [ZKM07b,TK14,PZM13, KMK15, TZW*16, HPLM16]. This makes this variation only an upper bound of the symmetrical PD g ; while sensible for motion planning, it is an arbitrary measure for two dynamic objects as the rotation of one of them is never taken into account.…”

Section: Background and Previous Workmentioning

confidence: 99%

“…[PZM13] have applied offline active learning methods to learn the contact space, however, their query time is still on the order of 1 ms. Additionally, their collision resolution method must separately recognize the contact‐areas, which in itself is slower than the combined query and gradient estimation time presented here. [KMK15,TZW*16] estimated the PD g using a samples database as well, though they are mainly aimed towards motion planning and do not address the computation of the additional data required for collision resolution. The query‐time of the latter is 0.03–3 ms. [HPLM16] is aimed towards dynamic simulations, but no collision resolution method is specified.…”

Section: Background and Previous Workmentioning

confidence: 99%

Fast Penetration Volume for Rigid Bodies

Nirel

Lischinski

2018

Computer Graphics Forum

View full text Add to dashboard Cite

Handling collisions among a large number of bodies can be a performance bottleneck in video games and many other real‐time applications. We present a new framework for detecting and resolving collisions using the penetration volume as an interpenetration measure. Given two non‐convex polyhedral bodies, a new sampling paradigm locates their near‐contact configurations in advance, and stores associated contact information in a compact database. At runtime, we retrieve a given configuration's nearest neighbors. By taking advantage of the penetration volume's continuity, cheap geometric methods can use the neighbors to estimate contact information as well as a translational gradient. This results in an extremely fast, geometry‐independent, and trivially parallelizable computation, which constitutes the first global volume‐based collision resolution. When processing multiple collisions simultaneously on a 4‐core processor, the average running cost is as low as 5 μs. Furthermore, no additional proximity or contact‐regions queries are required. These results are orders of magnitude faster than previous penetration volume approaches.

show abstract

Efficient global penetration depth computation for articulated models

Cited by 4 publications

References 32 publications

Active Learning of Neural Collision Handler for Complex 3D Mesh Deformations

Active Learning of Neural Collision Handler for Complex 3D Mesh Deformations

The manifold particle filter for state estimation on high-dimensional implicit manifolds

Fast Penetration Volume for Rigid Bodies

Contact Info

Product

Resources

About