AI Wheelchair Advances Toward Real-World Use

For individuals with severe disabilities, navigating tight spaces in a wheelchair often demonstrates a level of skill that surpasses current robotic capabilities. A significant focus of new research is determining whether AI-powered smart wheelchairs can effectively bridge this gap, and if such a goal is even desirable. Recent findings presented at a major conference highlight both the rapid progress and the complex challenges facing this field.

At the German Research Center for Artificial Intelligence, a team led by Christian Mandel and Serge Autexier has developed prototype sensor-equipped wheelchairs. These devices are engineered to maneuver through rooms filled with obstacles. The research, part of the REXASI-PRO project, integrated data from the wheelchair’s own sensors with external inputs, including drone-mounted cameras, to create a comprehensive safety system.

Mandel explains that their smart wheelchairs operate in two distinct modes. “Semiautonomous is the shared control system where the person sitting in the wheelchair uses the joystick to drive,” he says. “Fully autonomous is controlled by natural-language input. You say, ‘Please drive me to the coffee machine.'”

The experimental wheelchairs were outfitted with an array of technology, including lidars, a 3D camera, and an embedded computer. In autonomous mode, they utilize the open-source ROS2 Nav2 navigation system, interpreting spoken commands to create a path. Simultaneous localization and mapping (SLAM) and local obstacle-avoidance controllers then guide the chair, adjusting in real-time to avoid detected barriers after the user confirms the initial command.

However, the path to widespread adoption is not solely technical. Pooja Viswanathan, CEO of Braze Mobility, emphasizes that cost and accessibility are critical hurdles. “Funding systems are often not designed to support advanced add-on intelligence unless there is very clear evidence of value and safety,” she notes. She also points to the challenge of reliability in real-world conditions and the need for personalized solutions, as users have vastly different physical and cognitive needs.

Her company’s approach reflects a pragmatic, user-centered philosophy. Instead of building fully autonomous chairs, Braze develops blind-spot sensors that can be added to existing wheelchairs, providing obstacle alerts without removing user control. This design supports the rider rather than aiming to replace their judgment.

This perspective is echoed by biomedical engineer Louise Devinge, who cautions that increased autonomy brings greater system complexity. “The more sensing, computation, and autonomy you add,” she states, “the harder it becomes to ensure robust performance across the full range of real-world environments that wheelchair users encounter.” The immediate challenge, therefore, is not creating a replacement for the user, but engineering a more effective partnership between human and machine.

Looking ahead, Mandel believes mainstream smart wheelchairs could be a reality within a decade. Viswanathan sees projects like REXASI-PRO as vital for long-term innovation, pushing the boundaries of intelligent navigation and explainable AI, which is essential for user trust and safety.

Mandel’s own motivation is rooted in a lesson learned early in his career. After developing a system controlled by a head joystick, he observed that users with severe disabilities often navigated narrow passages with remarkable skill. “I realized, okay, there is this need for this technology,” he recalls, “but never underestimate what [wheelchair users] can do without it.” This balance between empowering assistance and preserving human capability remains the guiding principle for the field’s future.