A.D.Rs
Making autonomous delivery robots more ergonomic for people in wheelchairs.
Role
Designer/Strategist
Team
3 designers/strategists
Timeline
3 months
Deliverable
Solution and report
PROJECT OVERVIEW
The goal of the project was to make it more ergonomic for people in wheelchairs to lift what they'd ordered out of an autonomous delivery robot (ADR).
This was a quarter long project for IME 320: Human Factors and Technology. The class was split into teams and assigned a human factors related topic, and then tasked with identifying a human factors related problem in that space and proposing a solution. My team was assigned robotics, and we chose to focus on these ADRs because of their recent launch at Cal Poly.
PROBLEM
It was ergonomically difficult for a person in a wheelchair to lift the groceries out of the ADR.
Cal Poly had recently partnered with a company to bring ADRs to campus to help deliver food to students, so our team wanted to examine them for human factors related issues and found that people in wheelchairs had to bend over awkwardly in order to reach what they'd ordered, putting strain on their necks and shoulders.
SOLUTION
We designed a sliding mechanism and made the ADR height adjustable to reduce strain on the body.
BRAINSTORMING
After a lot of research on different kinds of robots, we thought something could be done to improve the design of ADRs. Something that sparked this interest was the company Starship launching their ADR food delivery service in Helsinki, Finland, where they sometimes get stuck in the snow. Because of this, we thought some kind of design improvement could be made to the wheels. However, we realized human factors wasn't really at the core of that problem, and eventually came across the ergonomic issues that arise from using them in a wheelchair.
DATA COLLECTION
To kick off the project, every group created a survey to start making improvements to their solution. Our form consisted of ten questions covering the respondent's experiencing using an ADR, what they thought could have been improved about it, and so on. We made sure to include a range of question types as well between open ended, multiple choice, and a scale of 1 to 10. Some of these questions included:
If a human factors engineer was making changes to an existing autonomous delivery robot, what changes might they make?
If you were designing an autonomous delivery robot for a person in a wheelchair, how would that change the design of the robot?
If you have used an autonomous delivery robot before, what were some things that could have been improved about the user experience? If not, please share some factors that would contribute to a positive user experience.
We received 21 responses to the survey, most of which came from our peers in the class as well as others we knew. Unfortunately, we were unable to find a person who actually used a wheelchair to participate in the survey, which will be addressed later in the case study.
After receiving these responses, we analyzed the results and began to draw conclusions. Here are some of the responses:
Our survey results were very insightful, while others were not as useful. For example, we asked participants about what would contribute to a positive and negative user experience when using a service like this. We received a lot of responses related to delivery time leading to a positive user exprerience. For these questions it was evident that the participants were not thinking from a human factors perspective; they were more concerned with convenience.
When asked about how to design an autonomous delivery robot specifically for users in wheelchairs, survey participants echoed our concern about reaching over, which was great to see. Importantly, this was before the question on the survey where they heard about our solution. 14 of the 21 participants on this question mentioned something related to accessibility, most of which pertained to the height of the robot. Some of which were related to reaching in as well.
Our last question prompted participants to share their thoughts on the solution we had put together, and while their feedback was a bit mixed, there were many positive comments like “very effective” and “this would be useful.”
ANTHROPOMETRIC STUDY
How could we accommodate the smallest and largest percentile?
Every team was then tasked with designing some kind of experiment or study that would provide them with data that would have to be implemented into the solution they would ultimately propose. We designed an anthropometric study where we identified key measurements related to total reach and subsequently took these measurements.
We had 18 people participate in our study, which consisted of taking four anthropometric measurements that we would use to determine the optimal height of the robot and length of the folding panel. The four measurements we took were arm length, torso length, hip to knee length, and total reach.
The averages for those metrics in the order they were just listed were 26.22in, 23.33in, 23.50, and 20.67. We were primarily concerned about total reach, however, so we made a bell curve for this metric. The standard deviation was 5.72, and the high and low extremes were 30.5in and 12.5in respectively. We also found a positive correlation between total reach and arm length, but a negative correlation between total reach and hip to knee length.
Anthropometrics was vital to our design because we needed data on the human body to determine which dimensions would make the most difference to the user. The main take-aways from our anthropometric study were the percentiles we needed to design for: the 5th percentile in total reach (9.23 in), the 5th percentile in arm length (22.89 in), the 5th percentile in torso length (20.91), and the 95th percentile in hip-to-knee length (20.67in).
IMPLEMENTING THE DATA
After some additional research, we were able to estimate how far the slider would need to extend and how how far it should be able to be raised to accommodate these percentiles:
Assumptions (kept conservative)
User is seated in a wheelchair and can’t bring the chair closer than their knee plane. Use 95th-percentile hip-to-knee = 20.67 in to set clearances so they don’t crash into taller users’ knees.
5th-percentile “total reach” = 9.23 in is treated as the maximum forward reach of the hand beyond the knee plane without strain.
User will keep a small safety standoff between the robot’s front face and their knee plane (recommend 1–2 in)
Present the leading edge of the groceries within the user’s reachable zone (≤ 9.23 in beyond the knee plane). Let:
S= safety standoff from the user’s knee plane (in)D= bin/deck depth measured from the robot’s front face to the leading edge of the groceries when retracted (in)R= user’s reach beyond knee plane (use 9.23 in from data)
Then the minimum extension of the slider is:
Extension ≥ S + (D − R)
Add ~2 in of design margin and round up to the next standard stroke.
Example
If you keep S = 2 in and the slider must clear a D = 18 in bin:
Required extension = 2 + (18 − 9.23) = 10.77 in
With margin → specify ~13 in stroke (or the next standard size, e.g., 14 in).
For vertical raise, use the short-arm users (5th percentile arm length = 22.89 in) to set the upper end. A comfortable “grip zone” for a seated pickup is roughly lap height up to ~10–12 in above lap without forcing trunk flexion or high shoulder elevation for short-armed users. Long-armed users don’t need extra height; they’re fine lower.
The 5th-percentile arm length is used because vertical access (how high you need to bring the groceries) is limited by how much vertical component a person can produce with their arm without uncomfortable shoulder elevation or trunk leaning. A conservative specification of up to +12 in of vertical raise was chosen because a short 5th-percentile arm (22.89 in) can produce about 10–11.5 in of comfortable vertical reach at realistic shoulder elevation angles.
FINAL PRODUCT
We designed our final solution in CAD and then 3d printed our prototype.
OUTCOMES
Here's what we accomplished.
We developed a solution to a human factors problem.
We were able to explain why our solution was feasible.
REFLECTIONS
Here's what I would do differently.
Interview the target user. It was difficult to design for a person in a wheelchair without speaking with one. We were able to learn a lot from our survey and our study but first hand expertise using a wheelchair would have certainly informed our design decisions.
Have more people participate in the study to more accurately gauge the percentiles.
Thanks for reading!