In STEM 2 we organize into teams to complete an engineering project. These projects are aimed at Assistive Technology, defined as assistive, adaptive, and rehabilitative devices for people with disabilities and the elderly. Throughout this three-month long project we engaged in workshops, research, and product development. For my project, I worked with Rachel, Charlotte, Garyth, Diego, David, and Aaron to develop an obstacle detection system for a wheelchair.

Problem Statement

An individual with a brain injury has difficulty navigating their home in a manual wheelchair due to impaired vision. Our target audience includes individuals with any degree of visual impairment who use a manual wheelchair.

Engineering Goal

The goal of this project is to design an attachable device for a wheelchair that allows a client with a visual impairment to navigate their home with increased safety and ease.

Design Approach

Designs- This project involved many steps and presentations including a Preliminary Design Review and Critical Design Review. Throughout the project our design iterations changed. At our first presentation, we explored sensors including infrared, LiDAR, and camera-based models. We also researched different options for feedback systems, including vibration pads, bracelets, and various buzzer modules. Eventually we settled on using a combination of sonar sensors for close-range and elevation detection, and LiDAR for far-range detection. We then began building mounts, configuring a Raspberry Pi and Arduinos, and assembling our vibration bracelets which provide feedback to the user. Many of our parts required 3D printing as well. In addition, we wanted to monitor the weight of the materials we used. Since it would be entirely mounted on a mobile wheelchair, it was important that the system was light. We also implemented our attachments in a way that would make the system adaptable to other wheelchairs, including Velcro on the sonar sensors and adjustable threaded rods on the LiDAR attachment base. Another important decision we made was to assemble two sonar sensors configured with their own Ardiunos and bracelets. This would allow us to deliver more detailed feedback to the user (for example right or left). In our testing we found that this helped to reduce the number of collisions.

Testing- Testing protocol was the same for each prototype. The MAMS Math room was utilized as a sort of obstacle course, with chairs tipped over and tables shuffled in a particular order that allowed just enough space for the wheelchair to navigate from the door to the whiteboard. Between designs, the layout of the room was identical, to avoid changing environments. The person testing the system was blindfolded to ensure that they could rely only on system-specific sensory cues to navigate their environment. They were given two minutes to navigate to their endpoint, and while they completed this task, the number of collisions and number of avoided collisions were counted and recorded.

Figure 1

Figure 1: Iterations of prototyping and number of collisions for each. Significant p values for each prototype in comparison with control. Complete system had highly significantly less collisions (**p < 0.001).

Future Extensions- To further optimize the device there are a variety of improvements which can be made. By placing a sonar sensor on the back of the wheelchair, the device would gain an additional field of vision. This would help make backing up more safe. Additionally, we would like to control the strength of the vibration to represent obstacles which are further or closer to the chair. For example, an object within a couple centimeters would produce a strong vibration, whereas one a foot or two away would vibrate less strongly. In an effort to further improve the safety of our user we would look to implement push notifications to a loved one, via text, to indicate that the user has potentially been in a collision. Finally, we would add a type of path generation which will sense the obstacles in the general area and guide the user through the obstacles in their path. This would help to reduce the amount of obstacles that the user collides with as well as reduce the time it would take for the user to change their path once an obstacle is detected.


Our final prototype included sonar sensors mounted on the foot rests, a LiDAR sensor mounted above the head of the user, and two bracelets (to direct left or right).

A photo at lake sunapee A photo at lake sunapee A photo at lake sunapee