ICRA 2020 Notes

Top 5 papers so far

• Learning Robust Control Policies for End-To-End Autonomous Driving from Data-Driven Simulation Alexander Amini, Igor Gilitschenski, Jacob Phillips, Julia Moseyko, Rohan Banerjee, Sertac Karaman, Daniela Ru https://ras.papercept.net/proceedings/ICRA20/3050.pdf https://ras.papercept.net/proceedings/ICRA20/3050.mp4 https://www.mit.edu/~amini/vista/

• Realistic extrapolation of visual driving scenes in a simulator

• ∇SLAM: Dense SLAM meets Automatic Differentiation Krishna Murthy Jatavallabhula1, Ganesh Iyer3, and Liam Paul https://ras.papercept.net/proceedings/ICRA20/2444.pdf http://montrealrobotics.ca/gradSLAM/ https://github.com/montrealrobotics/gradSLAM

  • RGBD
  • RGBD -> TSDF -> ICP -> Map Fusion
  • Levenberg Marquardt (smoother gradients)
  • Differentiable mapping (soft association)
  • negligible impact on accuracy
  • Soft association? Computational load?
  • How to extend it to arbitrary memory storage?

• Schmidt-EKF-based Visual-Inertial Moving Object Tracking Kevin Eckenhoff, Patrick Geneva, Nathaniel Merrill, and Guoquan Huang https://ras.papercept.net/proceedings/ICRA20/0294.pdf https://ras.papercept.net/proceedings/ICRA20/0294.mp4

  • Track dynamic object why estimating egomotion.
  • A sort of dynamic slam with one moving body
  • Model the target as a random walk and them apply EKF

• Graduated Non-Convexity for Robust Spatial Perception: From Non-Minimal Solvers to Global Outlier Rejection HENG YANG, Pasquale Antonante, Vasileios Tzoumas, Luca Carlone https://ras.papercept.net/proceedings/ICRA20/3735.pdf https://ras.papercept.net/proceedings/ICRA20/3735.mp4

• ρ(r(yᵢ, x)) Graduated non-convexity. Change ρ slowly from convex function to a robust non-convex function.

• Probabilistic Data Association via Mixture Models forRobust Semantic SLAM Kevin J. Doherty1, David P. Baxter1;�, Edward Schneeweiss1;2;�, and John J. Leonard (CSAIL) https://ras.papercept.net/proceedings/ICRA20/1656.pdf https://ras.papercept.net/proceedings/ICRA20/1656.mp4

  • Multiple hypothesis data association
  • Marginalizing out all possible hypothesis
  • Akin to max-mixtures Olson and Agarwal 2013
  • Add a binary data-association variable and the factor depends on the data-association variable.
  • GTSAM based optimization
  • Real time performance

Other papers

• Visual localization with Google Earth images for robust global pose estimation of UAVs B. Patel, T. D. Barfoot, and A. P. Schoellig in Proc. of the IEEE International Conference on Robotics and Automation (ICRA), 2020. Accepted.

• Metrically-Scaled Monocular SLAM Using Learned Scale Factors https://ras.papercept.net/proceedings/ICRA20/0017.pdf Greene, William Nicholas Massachusetts Institute of Technology Roy, Nicholas

  • combine depth prediction with metric unscaled slam Godard etal. [19],
  • Disparity estimator and predict D_l and D_r that map from I_l and I_r
  • Self supervised disparity estimator
  • With pyramids
  • Factor graph build from visual-inertial odometry

• Inertial-Only Optimization for Visual-Inertial Initialization https://ras.papercept.net/proceedings/ICRA20/0447.pdf

  • Focuses on initialization for VIO. (a) Vision only (b) Inertial only (c) Visual inertial.

• The empirical duality gap of constrained statistical learning https://arxiv.org/abs/2002.05183

  • Establishes the gap between parameteric approximation

• Hierarchical Quadtree Feature Optical Flow Tracking Based Sparse Pose-Graph Visual-Inertial SLAM https://ras.papercept.net/proceedings/ICRA20/0584.pdf

  • Focuses on optical flow feature tracking for VIO
  • by using hierarchical quadtree based searching of matching image patches.

• Robust 6D Object Pose Estimation by Learning RGB-D Features Meng Tian1, Liang Pan1, Marcelo H Ang Jr1and Gim Hee Lee2 https://ras.papercept.net/proceedings/ICRA20/0915.pdf

Autonomous driving papers

• Advanced BIT* (ABIT*): Sampling-Based Planning with Advanced Graph-Search Techniques Strub, Marlin Polo University of Oxford Gammell, Jonathan University of Oxford https://ras.papercept.net/conferences/conferences/ICRA20/proceedings/ICRA20_ContentListMedia_1.html RRT* Approximation from search BIT* Prioritizing efficiency Exploration vs Exploitation

• Deep merging Toyoto Motor Merging

• Robust Lane Detection with Binary Integer Optimization https://ras.papercept.net/conferences/conferences/ICRA20/proceedings/ICRA20_ContentListMedia_1.html Robust Lane Detection with Binary Integer Optimization Brandes, Kathleen Massachusetts Institute of Technology Wang, Allen Massachusetts Institute of Technology Shah, Rushina

Add constraints to the lane detection for continuity and against loops.

• A Synchronization Approach for Achieving Cooperative Adaptive Cruise Control Based Non-Stop Intersection Passing https://ras.papercept.net/conferences/conferences/ICRA20/proceedings/ICRA20_ContentListMedia_1.html

Award papers

• Active Reward Learning for Co-Robotic Vision Based Exploration inBandwidth Limited Environment

  • Directly update the reward model based on queries from the expert.

• Nonlinear Vector-Projection Control for Agile Fixed-Wing Unmanned Aerial Vehicles Juan Carlos Hernandez Ramirez, Meyer Nahon https://ras.papercept.net/proceedings/ICRA20/3732.pdf https://ras.papercept.net/proceedings/ICRA20/3732.mp4

• Metrically-Scaled Monocular SLAM Using Learned Scale Factors William Nicholas Greene, Nicholas Roy https://ras.papercept.net/proceedings/ICRA20/3734.pdf https://ras.papercept.net/proceedings/ICRA20/3734.mp4