Qiangqiang Huang

I am a PhD student at the Marine Robotics Group, part of CSAIL at MIT, where I work on inference algorithms for SLAM. My PhD advisor is Professor John Leonard. I received my B.E. and M.S. degrees from Tsinghua University.

My research lies at the intersection of robotics, Bayesian inference, computer vision, and machine learning. Specifically, I focus on developing full posterior inference algorithms that enable uncertainty-aware robotic perception. Applications of the algorithms include evaluating the uncertainty in localization and mapping, as well as supporting robots to plan how to reduce uncertainty for safe navigation.

Email  /  GitHub  /  Google Scholar  /  LinkedIn /  CV

Recent publications

GAPSLAM: Blending Gaussian Approximation and Particle Filters for Real-Time Non-Gaussian SLAM

A real-time SLAM solver that infers marginal posteriors of robot poses and landmark locations. We demonstrate the evolution of posteriors on realworld range-only SLAM and object-based SLAM datasets.

Optimizing Fiducial Marker Placement for Improved Visual Localization

An algorithmic solution to the problem of automatic marker placement within a scene. Specifically, given a predetermined set of markers and a scene model, we compute optimized marker positions within the scene that can improve accuracy in visual localization.

Incremental Non‑Gaussian Inference for SLAM Using Normalizing Flows

A solver that blends normalizing flows and the Bayes tree (aka. junction tree) to incrementally infer the joint posterior encountered in SLAM. Experiments were performed on range-only SLAM problems with data association ambiguity. This paper is an extension of our paper published in ICRA 2021.

Nested Sampling for Non-Gaussian Inference in SLAM Factor Graphs

A Monte Carlo approach that leverages nested sampling to generate high‑quality samples of posterior distributions at the expense of computation. These samples serve as reference solutions for validating other inference methods.

A Multi‑Hypothesis Approach to Pose Ambiguity in Object‑Based SLAM

This work presents a learned pose estimation network that provides multiple hypotheses about the 6D pose of an object, and formulates these hypotheses via mixture models in a SLAM backend.

Consensus-Informed Optimization Over Mixtures for Ambiguity-Aware Object SLAM

A real-time object-based SLAM system that is robust to symmetry- or occlusion-induced pose ambiguity from individual 6D object pose predictions.

NF‑iSAM: Incremental Smoothing and Mapping via Normalizing Flows

An algorithm that exploits the expressive power of neural networks, and trains normalizing flows to model and sample the joint posterior encountered in SLAM.

Design and source code from Leonid Keselman's Jekyll fork of Jon Barron's website