03.06.2026
Shared Control
A powered limb needs a way to work with the person using it. The person knows where they want to go and how they want to move. The controller has to turn the signals available to it into useful assistance, while responding to changes in speed, loading, and the environment.
We’re developing Genesis around shared control. The user provides movement intent, and ADA, our Adaptive Drive Architecture, handles more of the timing and low-level adjustments involved in delivering assistance. The balance between those roles has to change as the person moves. We want that to happen continuously, without making the user manage a sequence of manual mode changes.
Personalization is part of this work. People load a limb differently, move between phases of gait differently, and respond differently when conditions change. A controller needs information about those differences before it can usefully take on more of the work. We describe that progression in three stages: baseline, co-adaptation, and collaboration.
At baseline, the system observes the person’s gait and builds a profile of how they move. Predictability is the priority. The user is doing most of the work, while the sensing and control system gathers the information needed to support more personalized assistance.
In co-adaptation, the controller begins taking on more of the repeated adjustments involved in movement. The person remains the source of intent. The goal is to use what the system has learned to improve the timing and delivery of assistance across changing conditions.
In the collaboration stage, the controller takes on a larger share of the low-level actuation work, while the person remains the source of intent. The goal is to leave the person freer to focus on where and how they want to move. We use these stages to describe our approach to personalization and the behavior we want to evaluate as the system develops.
The engineering is closely coupled. Continuous sensing and low-latency control have to be supported by a reliable way to decide how user input and controller output interact. Personalization also needs limits so that adaptation does not make the system less predictable. Everyday conditions and unusual cases both belong in that validation work.
Motion data is useful because it lets us examine what the system actually did and where the response needs to change. We use that information to develop and evaluate subsequent controllers. The aim is a limb that can provide more assistance while requiring less conscious management. Shared control gives us a way to organize the work toward that experience.