We present a differentiable physics engine that can be integrated as a module in deep neural networks for end-to-end learning. As a result, structured physics knowledge can be embedded into larger systems, allowing them, for example, to match observations by performing precise simulations, while achieves high sample efficiency. Specifically, in this paper we demonstrate how to perform backpropagation analytically through a physical simulator defined via a linear complementarity problem. Unlike traditional finite difference methods, such gradients can be computed analytically, which allows for greater flexibility of the engine. Through experiments in diverse domains, we highlight the system's ability to learn physical parameters from data, efficiently match and simulate observed visual behavior, and readily enable control via gradient-based planning methods. Code for the engine and experiments is included with the paper.

The paper proposes to include an intermediate differentiable physics "engine" i.e. a novel parameterization of an intermediate layer in a neural network that respects the forward dynamics as governed by physics. The proposed method is closer to the popular "system identification" paradigm and involves learning the parameters of the engine via gradient-decent, optimizing the squared loss between observed and predicted frames. The paper shows results on some simple, interesting simulation domains and demonstrates the value of including such a module. Depending on the rebuttal for some of the questions below, I am happy to revise my ratings for this paper. Positives: - In general, the paper addresses an interesting and challenging question of incorporating more domain / structured knowledge into differentiable and learnable networks.

Designing motion control and planning algorithms for multilift systems remains challenging due to the complexities of dynamics, collision avoidance, actuator limits, and scalability. Existing methods that use optimization and distributed techniques effectively address these constraints and scalability issues. However, they often require substantial manual tuning, leading to suboptimal performance. This paper proposes Auto-Multilift, a novel framework that automates the tuning of model predictive controllers (MPCs) for multilift systems. We model the MPC cost functions with deep neural networks (DNNs), enabling fast online adaptation to various scenarios. We develop a distributed policy gradient algorithm to train these DNNs efficiently in a closed-loop manner. Central to our algorithm is distributed sensitivity propagation, which is built on fully exploiting the unique dynamic couplings within the multilift system. It parallelizes gradient computation across quadrotors and focuses on actual system state sensitivities relative to key MPC parameters. Extensive simulations demonstrate favorable scalability to a large number of quadrotors. Our method outperforms a state-of-the-art open-loop MPC tuning approach by effectively learning adaptive MPCs from trajectory tracking errors. It also excels in learning an adaptive reference for reconfiguring the system when traversing multiple narrow slots.

Differentiable physics is a powerful approach to learning and control problems that involve physical objects and environments. While notable progress has been made, the capabilities of differentiable physics solvers remain limited. We develop a scalable framework for differentiable physics that can support a large number of objects and their interactions. To accommodate objects with arbitrary geometry and topology, we adopt meshes as our representation and leverage the sparsity of contacts for scalable differentiable collision handling. Collisions are resolved in localized regions to minimize the number of optimization variables even when the number of simulated objects is high. We further accelerate implicit differentiation of optimization with nonlinear constraints. Experiments demonstrate that the presented framework requires up to two orders of magnitude less memory and computation in comparison to recent particle-based methods. We further validate the approach on inverse problems and control scenarios, where it outperforms derivative-free and model-free baselines by at least an order of magnitude.

We present a differentiable physics engine that can be integrated as a module in deep neural networks for end-to-end learning. As a result, structured physics knowledge can be embedded into larger systems, allowing them, for example, to match observations by performing precise simulations, while achieves high sample efficiency. Specifically, in this paper we demonstrate how to perform backpropagation analytically through a physical simulator defined via a linear complementarity problem. Unlike traditional finite difference methods, such gradients can be computed analytically, which allows for greater flexibility of the engine. Through experiments in diverse domains, we highlight the system's ability to learn physical parameters from data, efficiently match and simulate observed visual behavior, and readily enable control via gradient-based planning methods. Code for the engine and experiments is included with the paper.

Indian Institute of Technology Mandi in collaboration with Control Society is organising a five-day workshop on'Learning and Control' from 22 to 26 July. The aim of the workshop is to address the existing need for a sound analytical foundation for Machine Learning (ML) and Artificial Intelligence (AI) with Control Theory. The workshop is jointly sponsored by IIT Mandi, Control Society, and Council of Scientific and Industrial Research (CSIR) India. The workshop on'Learning and Control' is a platform to discuss current advances in the field of Machine Learning and Artificial Intelligence. The objective is to enhance the knowledge of participants who want to become researchers and expert users of Machine Learning and Control Methodologies, the workshop is designed with a focus on senior B. Tech students, research scholars and junior faculty from engineering institutes and colleges.

With a view to addressing the existing need for a sound analytical foundation for Machine Learning (ML) and Artificial Intelligence (AI), Indian Institute of Technology (IIT) Mandi is organising a workshop to train students and researchers on the issue. The workshop is being held with the collaboration of IIT Mandi, Control Society, and Council of Scientific and Industrial Research (CSIR) India from 22 to 26 July. The workshop on'Learning and Control' is a platform to discuss current advances in the field of Machine Learning and Artificial Intelligence and it is organised for the very first time. The objective is to enhance the knowledge of participants who want to become researchers and expert users of Machine Learning and Control Methodologies. The workshop is designed with a focus on senior B Tech students, research scholars and junior faculty from engineering institutes and colleges.