Alexander Luis Mitchell

Research Scientist & Robotics Engineer

I am a Robotics Engineer specializing in the intersection of Machine Learning and low-level control. Currently, I am working on whole-body planning and compliant control architectures for humanoid robots at Kinisi.

Previously, I was a Postdoc at the University of Oxford (Applied AI Lab) with Prof. Ingmar Posner. I completed my DPhil at the Oxford Robotics Institute under Prof. Posner and Prof. Ioannis Havoutis.

Dual-Arm Manipulation & Planning

My current research focuses on reactive assembly of tight-tolerance components. The video below demonstrates low-level compliant control integrated with high-level planning using energy-based models. The system handles both assistive and adversarial human interactions in real-time.

Impact & Outreach

Our platform, "Frank," serves as a testbed for research spanning elderly care, world models, and small-batch manufacturing. Recently, we demonstrated these capabilities to the Vice-Chancellor of the University of Oxford.

Locomotion & Representation Learning

During my DPhil, I explored how structured latent spaces can enable versatile robot locomotion. By learning a unified representation, we can generate multiple gaits and continuous transitions over uneven terrain.

Visualization of latent-space traversal for gait switching:

Latent Space Planner Visualization

Research Interests

Selected Publications

Gaitor: Learning a Unified Representation Across Gaits for Real-World Quadruped Locomotion
Conference of Robot Learning (CoRL), 2024
A. Mitchell, W. Merkt, A. Papatheodorou, I. Havoutis, I. Posner
Brain-like latent dynamics emerge in robot systems during walking and reaching
Nature: Scientific Reports, 2024
O. Parker Jones*, A. Mitchell*, J. Yamada*, et al.
VAE-Loco: Versatile Quadruped Locomotion by Learning a Disentangled Gait Representation
IEEE Transactions on Robotics (T-RO), 2023
A. Mitchell, W. Merkt, M. Geisert, et al.