Fundamental Challenges in Deep Learning for Stiff Contact Dynamics

Modeling, analysis and control of robot–object nonsmooth underactuated Lagrangian systems: A tutorial overview and perspectives

“…Friction models can be derived from data-driven algorithms [102]. An important point is to determine to which accuracy data-driven methods approximate the complex dynamics (1) so that it does not deteriorate too much the performance, to which type of task they apply ( [400] points out serious difficulties with too stiff contact, [397] incorporate unilateral constraints in their model-based reinforcement learning algorithm), and how much effort is necessary to implement them (data-driven methods being far from some kind of universal solution to all problems [389]). Koopman operators method, which consists of linearizing a finite-dimensional system by replacing it with an infinite-dimensional operator [401,402] has been applied to some hybrid systems [402,403].…”

Learning Object Manipulation With Under-Actuated Impulse Generator Arrays

“…Recent interest in model-based reinforcement learning has renewed research efforts to find good methods for learning world models [10]. Recently proposed Contact Nets have improved considerably the accuracy of predictive models with respect to earlier dynamical models based on standard neural networks [11], [12]. However, learning such models is quite complicated, and might also be an overkill for our control problem, where prediction of the entire future trajectory of the manipulated part is not really necessary, and predicting the stable resting state would suffice.…”

Bridging the Intention–Behavior Gap: Effect of Altruistic Motives on Developers’ Action towards Green Redevelopment of Industrial Brownfields

“…While earlier studies concentrated on learning these skills directly on physical robot systems, the focus has recently shifted to conducting most of the learning in simulations to address the training safety and efficiency. However, accurate simulation of complex contact dynamics is still a considerable challenge since simulation is sensitive to parameters like surface friction and material stiffness [9]. Moreover, task-specific geometries, such as irregular connectors, and tight peg-hole Fig.…”