This week I worked on expanding on the design of our project in preparation for the design presentation. Since we decided that Bluetooth and GPS would not be a viable solution for location tracking, I helped research other methods of location tracking, and our team decided to use UWB. From there, I looked into Apple’s Nearby Interaction framework which supports UWB and picked out our UWB module (DWM3001CTR13) that would be placed on the cart. The module and the iPhone allow for UWB connection which can be used to ensure that the shopping cart stays within range of the shopper.
I also researched how to integrate LiDAR and pathfinding for our cart. We decided to use a 2D 360 LiDAR sensor connected to a Raspberry Pi 5. From there, I looked into how we could get the information from that LiDAR sensor and combine it with the information from the UWB module to direct the movement of the cart. Robot Operating System (ROS) is a popular framework for building robot applications which we plan on using, but there may be issues with using it as it may not operate well on Macs. I read a research paper called “Development of Human Following Mobile Robot System Using Laser Range Scanner” which details how the group created a human following robot with LiDAR. In this paper, they mention the collision avoidance function they used, which outlines a collision detection area. If there is an obstacle in the collision detection area, the obstacle has its own circular collision avoidance area. The function then calculates the collision avoidance vector which is tangent to the circular collision avoidance area and in the same direction as the shopper’s direction. This vector is the direction in which the robot will travel. Since this is very similar to our project, we plan on using a similar function for our cart, integrating the coordinates of the shopper that we get from UWB.
We are currently on track for this week since most of the focus is on figuring out the design and design presentation. Although the function from the research article is a good option, there are more pathfinding algorithms that can be explored, so for next week, I plan on having a finalized pathfinding algorithm.