Enforcing the consensus between Trajectory Optimization and Policy Learning for precise robot control

Lidec, Quentin Le; Jallet, Wilson; Laptev, Ivan; Schmid, Cordelia; Carpentier, Justin

doi:10.1109/icra48891.2023.10160387

Cited by 1 publication

(1 citation statement)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Adopting a compliant contact force model has proven helpful [16] in overcoming the limitations faced by traditional approaches. The core idea of the continuous compliant contact model is a spring damping model in which the surface is regularly filled with parameters related to the surface material coefficient [17].…”

Section: Introductionmentioning

confidence: 99%

Variable Time-Step Physics Engine with Continuous Compliance Contact Model for Optimal Robotic Grinding Trajectory Planning

Zhou,

Pan,

Chen

et al. 2024

Sensors

View full text Add to dashboard Cite

In the transition from virtual environments to real-world applications, the role of physics engines is crucial for accurately emulating and representing systems. To address the prevalent issue of inaccurate simulations, this paper introduces a novel physics engine uniquely designed with a compliant contact model designed for robotic grinding. It features continuous and variable time-step simulations, emphasizing accurate contact force calculations during object collision. Firstly, the engine derives dynamic equations considering spring stiffness, damping coefficients, coefficients of restitution, and external forces. This facilitates the effective determination of dynamic parameters such as contact force, acceleration, velocity, and position throughout penetration processes continuously. Secondly, the approach utilizes effective inertia in developing the contact model, which is designed for multi-jointed robots through pose transformation. The proposed physics engine effectively captures energy conversion in scenarios with convex contact surface shapes through the application of spring dampers during collisions. Finally, the reliability of the contact solver in the simulation was verified through bouncing ball experiments and robotic grinding experiments under different coefficients of restitution. These experiments effectively recorded the continuous variations in parameters, such as contact force, verifying the integral stability of the system. In summary, this article advances physics engine technology beyond current geometrically constrained contact solutions, enhancing the accuracy of simulations and modeling in virtual environments. This is particularly significant in scenarios wherein there are constant changes in the outside world, such as robotic grinding tasks.

show abstract

Section: Introductionmentioning

confidence: 99%