Tristan Laidlow

Research Lead (SpatialAI), Boston Dynamics

Biography

I am the Research Lead for SpatialAI on the Atlas project at Boston Dynamics, where I take state-of-the-art perception algorithms and use them to enable new autonomous robot behaviors. In particular, my work is focused on combining semantic and geometric information (from both model-based and data-driven sources) so robots can understand and perform tasks in their environments. I am involved in the entire Atlas perception stack, from calibration and camera drivers to developing and maintaining cutting-edge machine learning models.

Previously, I was a research fellow in the Dyson Robotics Lab with Prof. Andrew Davison at Imperial College London. During my PhD, I worked under the supervision of Prof. Dr. Stefan Leutenegger on making real-time dense visual SLAM more robust through the fusion of inertial data and learned depth priors. I completed a bachelor’s degree in aerospace engineering from the University of Toronto, where my thesis project was supervised by Prof. Dr. Angela Schoellig.

Videos

Perception and Adaptability | Inside the Lab with Atlas

I was interviewed for this behind-the-scenes video at Boston Dynamics. It showcases the work our team has been up to for the last year or so, demonstrating how our machine learning and real-time perception systems enable Atlas to perform accurate and robust manipulation tasks in a factory-like environment.

Atlas Goes Hands On

A technical demonstration of the type of robot behaviors our team is working on. After receiving an input list of bin locations, Atlas autonomously moves engine covers between supplier containers and a mobile sequencing dolly in the specified order. Atlas is using a learned vision model I designed to detect and localize the fixtures and individual bins in the environment.

HD Atlas Manipulates

This video highlights some of the very early object detection and localization work I was involved with on HD Atlas. Most of the techniques we were using at this stage were very basic, but taught us a lot about how to manipulate heavy and difficult-to-perceive objects.

Atlas Gets a Grip

The very first Boston Dynamics video I got to help make (I mostly worked on the object localization)! It was very stressful but also a lot of fun!

Publications

iMODE reconstructions

iMODE: Real-Time Incremental Monocular Dense Mapping Using Neural Field

Hidenobu Matsuki, Edgar Sucar, Tristan Laidlow, Kentaro Wada, Raluca Scona, Andrew J. Davison

ICRA 2023 (Best Navigation Paper Award Finalist)

BodySLAM pipeline

BodySLAM: Joint Camera Localisation, Mapping, and Human Motion Tracking

Dorian Henning, Tristan Laidlow and Stefan Leutenegger

ECCV 2022

quadrics-based reconstruction

Simultaneous Localisation and Mapping with Quadric Surfaces

Tristan Laidlow and Andrew J. Davison

3DV 2022

Robot arm using C2F

Coarse-to-Fine Q-Attention: Efficient Learning for Visual Robotic Manipulation via Discretisation

Stephen James, Kentaro Wada, Tristan Laidlow, and Andrew J. Davison

CVPR 2022 (Oral)

iLabel reconstruction

iLabel: Interactive Neural Scene Labelling

Shuaifeng Zhi*, Edgar Sucar*, Andre Mouton, Iain Haughton, Tristan Laidlow, and Andrew J. Davison

(* denotes equal contribution)

RA-L 2022

SemanticNeRF pipeline

In-Place Scene Labelling and Understanding with Implicit Scene Representation

Shuaifeng Zhi, Tristan Laidlow, Stefan Leutenegger, and Andrew J. Davison

ICCV 2021 (Oral)

SIMstack

SIMstack: A Generative Shape and Instance Model for Unordered Object Stacks

Zoe Landgraf, Raluca Scona, Tristan Laidlow, Stephen James, Stefan Leutenegger, and Andrew J. Davison

ICCV 2021

Learned and geometric probability distributions

Towards the Probabilistic Fusion of Learned Priors into Standard Pipelines for 3D Reconstruction

Tristan Laidlow, Jan Czarnowski, Andrea Nicastro, Ronald Clark, and Stefan Leutenegger

ICRA 2020

DeepFactors

DeepFactors: Real-Time Probabilistic Dense Monocular SLAM

Jan Czarnowski, Tristan Laidlow, Ronald Clark, and Andrew J. Davison

RA-L 2020

DeepFusion

DeepFusion: Real-Time Dense 3D Reconstruction for Monocular SLAM using Single-View Depth and Gradient Predictions

Tristan Laidlow, Jan Czarnowski, and Stefan Leutenegger

ICRA 2019

Learned Mesh Reconstruction

Learning Meshes for Dense Visual SLAM

Michael Bloesch, Tristan Laidlow, Ronald Clark, Stefan Leutenegger, and Andrew J. Davison

ICCV 2019

RGB-D-inertial reconstruction

Dense RGB-D-Inertial SLAM with Map Deformations

Tristan Laidlow, Michael Bloesch, Wenbin Li, and Stefan Leutenegger

IROS 2017

primer

A Primer on the Differential Calculus of 3D Orientations

Michael Bloesch, Hannes Sommer, Tristan Laidlow, Michael Burri, Gabriel Nuetzi, Peter Fankhauser, Dario Bellicoso, Christian Gehring, Stefan Leutenegger, Marco Hutter, and Roland Siegwart

arXiv preprint 2016