Roll with the Changes

The technologies underpinning modern self-driving vehicles are progressing rapidly, but even still, a lot of people do not yet feel comfortable giving full control to their car. Will it detect a pedestrian that unexpectedly appears in the middle of the road? Or that other vehicle that is weaving in and out of traffic? We have to have a lot of trust in these systems to believe that they will keep us and others safe in such situations.

But then there are the safety features that keep us between the lines, alert us of potential collisions before they happen, and so on. These types of features can give us peace of mind without completely handing over the wheel for the duration of the drive. People tend to trust these features far more, and find them to be very convenient for everything from monotonous daily commutes to long trips.

Researchers at the University of Michigan believe that these types of technologies might also benefit individuals that rely on wheelchairs for mobility. Traditionally, wheelchairs are either fully manual or fully autonomous. Why is there no middle ground for those that want some help getting around, but do not want to give full control to a machine?

To address this gap, the research team has developed CoNav, a smart wheelchair that merges manual and autonomous control. The CoNav chair is designed to assist individuals with mobility impairments by allowing them to control the wheelchair while also benefiting from an intelligent navigation system.

Most existing wheelchair designs have significant drawbacks. Manual wheelchairs require constant user input, which can be difficult for individuals with limited motor function. Fully autonomous wheelchairs, on the other hand, struggle in dynamic environments and often fail to win the trust of their users.

CoNav offers a shared control approach, where users can steer the wheelchair via a joystick while the system actively assists in navigation, avoiding obstacles and improving efficiency. If a user prefers more control, the chair adapts, prioritizing their input. When users are less active in controlling the wheelchair, the system takes over to ensure smooth and safe movement.

The team’s system is built on top of a commercially available Quantum Q6 Edge 2.0 powered wheelchair, modified with a range of sensors and an advanced Robot Operating System-based framework. The chair is equipped with LiDAR sensors for real-time obstacle detection and mapping, a stereo camera for visual perception of the environment, an inertial measurement unit to track movement and stability, and encoders to precisely measure wheel rotations for accurate positioning.

At the core of CoNav’s navigation system is a Model Predictive Control algorithm, which continuously predicts the best movements based on user input and environmental conditions. The system further uses Simultaneous Localization and Mapping to create a map of the surroundings, allowing the wheelchair to navigate safely and efficiently.

During real-world testing, CoNav demonstrated some advantages over both manual and autonomous wheelchairs. It significantly reduced abrupt movements, improved navigation in tight spaces, and minimized collisions with obstacles in a series of experiments. Users reported feeling more comfortable and in control while also appreciating the wheelchair’s assistance in tricky situations.

Looking ahead, the research team aims to refine the chair’s capabilities, incorporating socially aware navigation for crowded spaces and multimodal user input such as eye-tracking and EEG signals. These enhancements would expand CoNav’s usability for individuals with varying mobility impairments.