Dissipative neural networks improve mechanical system predictions and understanding
Dissipative Lagrangian Neural Networks for the Interpretable Modelling of Passive Mechanical Systems
Machine LearningComputational Engineering, Finance, and Science
Summary
Some machines and mechanical systems have parts that slow down movement, like friction, which can be tricky to predict. This work introduces a new kind of neural network that learns to separate out these slowing forces from normal movement forces using data. The researchers tested their method on a swinging double pendulum and a robot arm, showing it predicts movements better and can explain itself with simple formulas. This helps when there isn’t much data available and makes the models easier to understand and use.
What this means in practice
- •For industrial robot programmers: Create predictive models of robot arm movement that balance accuracy with interpretability and speed for real-time control.
- •For robotics algorithm engineers: Develop interpretable models for mechanical systems that generalize well with limited data, improving simulation and control.
Authors
Rob Seabourne, Alice Cicirello
Abstract
In this paper, we propose the Dissipative Lagrangian Neural Network (dLNN), a continuous-time system identification technique that separately learns the conservative and dissipative dynamics of a mechanical system directly from data. A key contribution is a novel parametrisation of dissipative forces that (i) is structurally constrained to be dissipative, enforcing passivity of the learned model, (ii) cannot absorb conservative mechanics, and (iii) retains sufficient expressivity to represent a broad family of state-dependent damping and regularised friction forces. The model is validated on prediction tasks on two real-world mechanical systems: trajectory prediction of a chaotic double pendulum and inverse-dynamics modelling of a six-degree-of-freedom industrial robot. Across these tasks, the dLNN demonstrates improved predictive performance over existing Lagrangian architectures while remaining competitive with a black-box neural model. The imposed physical structure also improves generalisation to unseen states, with the benefits becoming particularly pronounced when training data is limited. Finally, we introduce a symbolic distillation procedure that extracts closed-form expressions from the learned physical components. Applied to the industrial robot, this produces interpretable symbolic models that retain high predictive accuracy while reducing the computational overhead associated with evaluating neural Lagrangian models.