Christopher Iliffe Sprague

Postdoctoral researcher at SciLifeLab

prof_pic2.jpg

Robotics, Perception, and Learning

Teknikringen 14

Stockholm, Sweden, 114 23

Research

My research sits at the crossroads of learning-based and model-based methods. On one hand, I aim to optimise the efficacy of model-based methods using learning techniques. On the other hand, I work to ensure the reliability of learning-based methods through model-based approaches.

My core areas of interest include robotics, physical sciences, and life sciences. These fields offer a wealth of inductive biases that can enhance learning-based methods, especially in low-data environments.

Currently, I’m expanding my research horizons as a postdoctoral researcher at SciLifeLab, in collaboration with Prof. Hossein Azizpour and Prof. Arne Elofsson. I’m exploring the potential of learning-based methods in modelling molecular interactions. My areas of interest in this domain include protein-protein interactions, drug discovery, and protein folding. Despite the scarcity of data in this field, it boasts a rich supply of geometric and physical principles that can be used to enhance learning-based models.

Prior to my current role, I pursued my PhD at the KTH Royal Institute of Technology in Stockholm, Sweden, in collaboration with Prof. Petter Ögren. My doctoral research focused on crafting efficient and trustworthy AI solutions for critical robotic systems, using behaviour trees. This work underscored the importance of efficiency and safety while exploring the potential of learning-based methods to achieve these goals, all within a framework of model-based reliability.

Before embarking on my PhD journey, I obtained both my Master’s and Bachelor’s degrees at Rensselaer Polytechnic Institute in Troy, NY, USA, specialising in aerospace engineering.

News

Latest posts

Feb 9, 2024 Stable Flow Matching

Selected publications

  1. Stable Autonomous Flow Matching
    Christopher Iliffe Sprague, Arne Elofsson, and Hossein Azizpour
    In submission, 2024
  2. Adding neural network controllers to behavior trees without destroying performance guarantees
    Christopher Iliffe Sprague, and Petter Ögren
    In 2022 IEEE 61st Conference on Decision and Control (CDC), 2022
  3. PointNetKL: Deep inference for GICP covariance estimation in bathymetric SLAM
    Ignacio Torroba, Christopher Iliffe Sprague, Nils Bore, and John Folkesson
    IEEE Robotics and Automation Letters, 2020
  4. Learning dynamic-objective policies from a class of optimal trajectories
    Christopher Iliffe Sprague, Dario Izzo, and Petter Ögren
    In 2020 59th IEEE Conference on Decision and Control (CDC), 2020
  5. Continuous-time behavior trees as discontinuous dynamical systems
    Christopher Iliffe Sprague, and Petter Ögren
    IEEE Control Systems Letters, 2021
  6. An Extended Convergence Result for Behavior Tree Controllers
    Christopher Iliffe Sprague, and Petter Ögren
    In submission, 2023