• Corpus ID: 209444192

Generating Robust Supervision for Learning-Based Visual Navigation Using Hamilton-Jacobi Reachability

@article{Li2020GeneratingRS,
  title={Generating Robust Supervision for Learning-Based Visual Navigation Using Hamilton-Jacobi Reachability},
  author={Anjian Li and Somil Bansal and Georgios Giovanis and Varun Tolani and Claire J. Tomlin and Mo Chen},
  journal={ArXiv},
  year={2020},
  volume={abs/1912.10120}
}
In Bansal et al. (2019), a novel visual navigation framework that combines learning-based and model-based approaches has been proposed. Specifically, a Convolutional Neural Network (CNN) predicts a waypoint that is used by the dynamics model for planning and tracking a trajectory to the waypoint. However, the CNN inevitably makes prediction errors which often lead to collisions in cluttered and tight spaces. In this paper, we present a novel Hamilton-Jacobi (HJ) reachability-based method to… 

Figures and Tables from this paper

LBGP: Learning Based Goal Planning for Autonomous Following in Front
TLDR
This paper investigates a hybrid solution which combines deep reinforcement learning (RL) and classical trajectory planning for the "following in front" application, and demonstrates zero-shot transfer of the trained policy from simulation to the real world.
Memory-Augmented System Identification With Finite-Time Convergence
This letter presents a memory-augmented system identifier with finite-time convergence for continuous-time uncertain nonlinear systems. A memory of events with significant effect on the performance
Sim-to-Lab-to-Real: Safe Reinforcement Learning with Shielding and Generalization Guarantees
TLDR
This paper applies a dual policy setup where a performance policy is training using the cumulative task reward and a backup policy is trained by solving the reach-avoid Bellman Equation based on Hamilton-Jacobi reachability analysis to improve safety.
SympOCnet: Solving optimal control problems with applications to high-dimensional multi-agent path planning problems
TLDR
A novel neural network method called SympOCnet is proposed that applies the Symplectic network to solve high-dimensional optimal control problems with state constraints and can solve a problem with more than 500 dimensions in 1.5 hours on a single GPU.
A Multimodal and Hybrid Framework for Human Navigational Intent Inference
TLDR
Experimental results reveal that the proposed learning-based prediction module outperforms various baselines especially when a relatively small amount of data is available, and the performance of the framework on a newly collected dataset called SFU-Store-Nav is evaluated.
Failing with Grace: Learning Neural Network Controllers that are Boundedly Unsafe
TLDR
This work considers the problem of learning a feed-forward neural network controller to safely steer an arbitrarily shaped planar robot in a compact and obstacle-occluded workspace and proposes an alternative approach that lifts such strong assumptions on the data that are hard to satisfy in practice and instead allows for graceful safety violations.
Hopf-type representation formulas and efficient algorithms for certain high-dimensional optimal control problems
TLDR
This paper provides the analytical solution for a class of optimal control problems whose running costs consist of a quadratic on the control variable and a convex, non-negative, piecewise affine function on the state variable and proposes efficient numerical algorithms based on the Hopf-type representation formula, convex optimization algorithms, and min-plus techniques.
Investigating Neural Network Architectures, Techniques, and Datasets for Autonomous Navigation in Simulation
TLDR
The results presented here show that there can often be a loose relationship between validation metrics and performance, and although trained models perform well in an environment matching the distribution of the training dataset, they fail when the simulation environment is altered in a seemingly insignificant manner.
Lax-Oleinik-type formulas and efficient algorithms for certain high-dimensional optimal control problems
TLDR
Using existing optimization methods and the min-plus technique, the numerical methods are extended to address more general classes of convex and nonconvex initial costs and are shown, in some cases, to scale linearly with the state dimension, and thus, to overcome the curse of dimensionality.
...
1
2
...

References

SHOWING 1-10 OF 34 REFERENCES
Combining Optimal Control and Learning for Visual Navigation in Novel Environments
TLDR
This work coupling model-based control with learning-based perception produces a series of waypoints that guide the robot to the goal via a collision-free path and demonstrates that the proposed approach can reach goal locations more reliably and efficiently in novel environments as compared to purely geometric mapping-based or end-to-end learning- based alternatives.
Target-driven visual navigation in indoor scenes using deep reinforcement learning
TLDR
This paper proposes an actor-critic model whose policy is a function of the goal as well as the current state, which allows better generalization and proposes the AI2-THOR framework, which provides an environment with high-quality 3D scenes and a physics engine.
Neural Autonomous Navigation with Riemannian Motion Policy
TLDR
This work designs a deep neural network to predict control point RMPs of the vehicle from visual images, from which the optimal control commands can be computed analytically, and shows that the network trained in the Gibson environment can be used for indoor obstacle avoidance and navigation on a real RC car.
An Efficient Reachability-Based Framework for Provably Safe Autonomous Navigation in Unknown Environments
TLDR
This work proposes a novel, real-time safety analysis method based on Hamilton-Jacobi reachability that provides strong safety guarantees despite the unknown parts of the environment and demonstrates the approach in simulation and in hardware to provide safety guarantees around a state-of-the-art vision-based, learning-based planner.
ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst
TLDR
The ChauffeurNet model can handle complex situations in simulation, and the perturbations then provide an important signal for these losses and lead to robustness of the learned model.
Cognitive Mapping and Planning for Visual Navigation
TLDR
The Cognitive Mapper and Planner is based on a unified joint architecture for mapping and planning, such that the mapping is driven by the needs of the task, and a spatial memory with the ability to plan given an incomplete set of observations about the world.
Deep Drone Racing: Learning Agile Flight in Dynamic Environments
TLDR
This approach combines a convolutional neural network (CNN) with a state-of-the-art path-planning and control system and generates a short, minimum-jerk trajectory segment and corresponding motor commands to reach the desired goal.
Bayesian Learning for Safe High-Speed Navigation in Unknown Environments
TLDR
By using a Bayesian non-parametric learning algorithm that encodes formal safety constraints as a prior over collision probabilities, a planner for high-speed navigation in unknown environments seamlessly reverts to safe behavior when it encounters a novel environment for which it has no relevant training data.
Beauty and the Beast: Optimal Methods Meet Learning for Drone Racing
TLDR
The presented approach was used to win the IROS 2018 Autonomous Drone Race Competition, outracing the second-placing team by a factor of two and demonstrating agile and robust flight through complex and diverse previously-unseen race tracks.
Learning to fly by crashing
TLDR
This paper builds a drone whose sole purpose is to crash into objects: it samples naive trajectories and crashes into random objects to create one of the biggest UAV crash dataset.
...
1
2
3
4
...