Disney Research is a privately owned entity, headquartered in the US, and operates under Disney Experiences. Founded in 2008, employing approximately 100 individuals, the company offers research services as a corporate network of scientific and technological research laboratories supporting the multi-faceted business segments of The Walt Disney Company. It operates primary facilities in Los Angeles, United States, and Zurich, Switzerland (including specialized arms such as DisneyResearch|Studios), alongside an academic partnership with ETH Zurich. Functioning as an internal innovation hub rather than a consumer-facing media vendor, it tasks international researchers with developing early-stage scientific concepts to enhance corporate assets. Its core areas of investigation encompass computer vision, robotics, machine learning, human-computer interaction, visual computing, and immersive technologies, generating proprietary hardware and software models—such as adaptive character controls, robotic motion diffusion frameworks, and omnidirectional locomotion surfaces—that are integrated across Walt Disney Studios productions and Disney Experiences theme park attractions.
On May 13, 2026, Disney Research introduced CoCo-InEKF, a differentiable Invariant Extended Kalman Filter utilizing a neural module to predict contact velocity covariances, enabling robust state estimation in dynamic contact-rich scenarios. This method is trained end-to-end via backpropagation through time and achieves improved performance over classical and other learning-based approaches. Previously, on May 12, Disney Research shared a method to improve robot movement accuracy, proposing a bilevel optimization framework that jointly adapts human kinematic reference motions to a robot's morphology while training a tracking policy using reinforcement learning, thereby addressing physical inconsistencies like foot sliding or self-collisions. This follows the May 10 demonstration of "ReActor," a system focused on transferring human motion to robots with diverse body shapes and movement capabilities, aiming to significantly enhance robotic movement. On May 6, the research had formally introduced the "ReActor" bilevel optimization framework, which bridges the embodiment gap between human motion and varied robotic morphologies by jointly adapting retargeting parameters and learning an optimal tracking policy using reinforcement learning. Earlier, on March 17, Disney Research announced a collaboration with NVIDIA to build physical Olaf robots, with NVIDIA providing the simulation platform and AI hardware to train the robots' movements, balancing, and interactions in a virtual environment before deployment.
Subscribe for full access to Disney Research's products in full detail
Subscribe for full access to Disney Research's revenue in full detail
Subscribe for full access to Disney Research's products in full detail 
