• Corpus ID: 235436267

Causal Navigation by Continuous-time Neural Networks

@inproceedings{Vorbach2021CausalNB,
  title={Causal Navigation by Continuous-time Neural Networks},
  author={Charles J. Vorbach and Ramin M. Hasani and Alexander Amini and Mathias Lechner and Daniela Rus},
  booktitle={Neural Information Processing Systems},
  year={2021}
}
Imitation learning enables high-fidelity, vision-based learning of policies within rich, photorealistic environments. However, such techniques often rely on traditional discrete-time neural models and face difficulties in generalizing to domain shifts by failing to account for the causal relationships between the agent and the environment. In this paper, we propose a theoretical and experimental framework for learning causal representations using continuous-time neural networks, specifically… 

Figures and Tables from this paper

Latent Imagination Facilitates Zero-Shot Transfer in Autonomous Racing

This paper investigates how model-based agents capable of learning in imagination substantially outperform model-free agents with respect to performance, sample efficiency, successful task completion, and generalization in real-world autonomous vehicle control tasks, where advanced model- free deep RL algorithms fail.

Are All Vision Models Created Equal? A Study of the Open-Loop to Closed-Loop Causality Gap

The results imply that the causality gap can be solved in situation one with the proposed training guideline with any modern network architecture, whereas achieving out-of-distribution generalization requires further investigations, for instance, on data diversity rather than the model architecture.

Closed-form continuous-time neural networks

It is shown that it is possible to closely approximate the interaction between neurons and synapses—the building blocks of natural and artificial neural networks—constructed by liquid time-constant networks efficiently in closed form and obtain models that are between one and five orders of magnitude faster in training and inference compared with differential equation-based counterparts.

Closed-form Continuous-Depth Models

This paper presents a new family of models, termed Closed-form Continuous-depth (CfC) networks, that are simple to describe and at least one order of magnitude faster while exhibiting equally strong modeling abilities compared to their ODE-based counterparts.

Interpreting Neural Policies with Disentangled Tree Representations

A new algorithm is designed that programmatically extracts tree representations from compact neural policies, in the form of a set of logic programs grounded by world state, that allows for interpretability metrics that measure disentanglement of learned neural dynamics from a concentration of decisions, mutual information and modularity perspectives.

Closed-form Continuous-time Neural Models

It is shown it is possible to closely approximate the interaction between neurons and synapses – the building blocks of natural and artificial neural networks – constructed by liquid time-constant networks (LTCs) efficiently in closed-form.

Sparse Flows: Pruning Continuous-depth Models

This work designs a framework to decipher the internal dynamics of these continuous depth models by pruning their network architectures, and empirical results suggest that pruning improves generalization for neural ODEs in generative modeling.

BarrierNet: A Safety-Guaranteed Layer for Neural Networks

These novel safety layers, termed a BarrierNet, can be used in conjunction with any neural network-based controller, and can be trained by gradient descent, which allows the safety constraints of a neural controller be adaptable to changing environments.

Social ODE: Multi-agent Trajectory Forecasting with Neural Ordinary Differential Equations

The Social ODE approach compares favor-ably with state-of-the-art methods, and more importantly, can successfully avoid sudden obstacles and effectively control the motion of the agent, while previous methods often fail in such cases.

Entangled Residual Mappings

While entangled mappings can preserve the iterative refinement of features across various deep models, they influence the representation learning process in convolutional networks differently than attention-based models and recurrent neural networks.

References

SHOWING 1-10 OF 104 REFERENCES

Learning Exploration Policies for Navigation

This work proposes a learning-based approach and finds that the use of policies with spatial memory that are bootstrapped with imitation learning and finally finetuned with coverage rewards derived purely from on-board sensors can be effective at exploring novel environments.

Conditional Affordance Learning for Driving in Urban Environments

This work proposes a direct perception approach which maps video input to intermediate representations suitable for autonomous navigation in complex urban environments given high-level directional inputs, and is the first to handle traffic lights and speed signs by using image-level labels only.

One-Shot Hierarchical Imitation Learning of Compound Visuomotor Tasks

This work proposes a method that learns both how to learn primitive behaviors from video demonstrations and how to dynamically compose these behaviors to perform multi-stage tasks by "watching" a human demonstrator.

Model-based versus Model-free Deep Reinforcement Learning for Autonomous Racing Cars

It is shown that model-based agents capable of learning in imagination, substantially outperform model-free agents with respect to performance, sample efficiency, successful task completion, and generalization, and that the generalization ability of model- based agents strongly depends on the observationmodel choice.

Deep Imitative Models for Flexible Inference, Planning, and Control

This paper proposes Imitative Models, probabilistic predictive models of desirable behavior able to plan interpretable expert-like trajectories to achieve specified goals, and derives families of flexible goal objectives that can be used to successfully direct behavior.

Gershgorin Loss Stabilizes the Recurrent Neural Network Compartment of an End-to-end Robot Learning Scheme

A new regularization loss component is introduced together with a learning algorithm that improves the stability of the learned autonomous system, by forcing the eigenvalues of the internal state updates of an LDS to be negative reals.

Learning by Cheating

This work shows that this challenging learning problem can be simplified by decomposing it into two stages and uses the presented approach to train a vision-based autonomous driving system that substantially outperforms the state of the art on the CARLA benchmark and the recent NoCrash benchmark.

Learning to Control PDEs with Differentiable Physics

It is shown that by using a differentiable PDE solver in conjunction with a novel predictor-corrector scheme, this work can train neural networks to understand and control complex nonlinear physical systems over long time frames.

Lipschitz Recurrent Neural Networks

This work proposes a recurrent unit that describes the hidden state's evolution with two parts: a well-understood linear component plus a Lipschitz nonlinearity, which is more robust with respect to input and parameter perturbations as compared to other continuous-time RNNs.

Closed-form Continuous-Depth Models

This paper presents a new family of models, termed Closed-form Continuous-depth (CfC) networks, that are simple to describe and at least one order of magnitude faster while exhibiting equally strong modeling abilities compared to their ODE-based counterparts.
...