Tuesday, March 14, 2006

RoboCart: Grocery shopping solution for the visually impaired

U0204714 Chan Hongjiang Human-robot interaction (HRI) has always been an exciting field of study in the area of assistive robotics; however, much of it has been based on the assumption that the human involved has visual capabilities of identifying and making use of the robot. In 2004, researchers from the Computer Science Assistive Technology Laboratory of the Department of Computer Science of Utah State University launched a project to build a robotic shopping cart for the visually impaired. The objective was to allow the visually impaired to be able to navigate and shop for grocery by themselves, with the aid of a robot. The RoboCart is designed to navigate to any location in the supermarket desired by the user. The RoboCart project takes into many considerations, most importantly of which is the ergonomical aspect of the human-robot interaction process for the visually impaired. A special handlebar with a keypad allows the user to input instructions to the shopping cart and hold on to the shopping cart for guidance with one hand only while leaving the other hand free for a guide dog or a white cane. Audio guidance is also provided to the user to pre-inform him of the direction the shopping cart is headed to next. This elegant solution was achieved only through much trial and error with various possible solutions like speech recognition, guide leash instead of a static handle, and dynamic Braille displays for output. Another area of research is how the RoboCart finds its way around the supermarket. Various solutions have been implemented for other robotic applications, including GPS, visual mapping through onboard/environment cameras, mapping through active beacons in intelligent space and so on. For the RoboCart project, the choice was a combination of position identification through RFID and line-following along the supermarket aisles. Line following is simple to implement and recent advancement in RFID has made it a very affordable technology for mass implementation. As the consumer robotics market advances, we can expect a boom in the number and types of assistive robots to be developed and marketed, catering to humans from all walks of life, for all walks of life :) References: http://cc.usu.edu/~cpg/pubs/iros05.pdf http://cc.usu.edu/~cpg/pubs/hri2006_ergo.pdf


Medical said...

U0204550 Wong Liang Mian

My guess is that some customization maybe possible with the robot. A blind person could have his/her ID and information stored in a memory device (ie: a mmc card) This device could then be inserted into the robot.

Most people are creatures of habit, as such the robot would be able to record what the person usually buys and plot the most efficient path to the items needed. The robot would also be able to record the person's ideal moving speed and adjust its own moving speed accordingly.

This robot, if combined with a robotic arm, could become a robot shopper as well. People would send their orders to the supermarket. The supermarket would download the details into the robot and let the robot acquire the goods within the supermarket. The supermarket would then pack the goods and send it to the customer.

Home said...

u0204438 Huang Shichao Alvin

I guess that the robot will also need to have some way to avoid obstacles in its path, because as we all know, even sighted people tend to run into each other sometimes. Also, it can also be envisioned that if this is widely implemented, it is likely that sighted people will also want to make use of it.

This will likely result in each supermarket having a substantial amount of these robots moving around in the supermarket. Thus, there will be a need for both simple obstacle avoidance strategies, such as IR sensors to detect other people and carts moving in front of it, as well as more complex path planning by an overall system to avoid congestion along popular routes.

It will be interesting, however, to see how these robots deal with the human tendency to make sharp turns, change their minds, etc.

Assistive said...

u0204714 Chan Hongjiang

The RoboCart does have obstacle avoidance capabilities, which I have failed to mention in my post, sorry for the oversight.

As for how it deals with the human tendencies to make sharp turns etc, actually that is a problem encountered, but in the reverse direction. As most of the time, the RoboCart will be the one to lead the user, hence it does not suffer from confusing instructions from the user. However, some blind testers have commented that when the RoboCart decides to make a U-turn, problems are encountered, as it is hard for the user to follow the lead of the RoboCart and use their walking sticks at the same time when the turning radius is tight. The turning radius is limited by the size of the walkway in the supermarket.

For the robotic arm suggestion, that is a good one as the RoboCart can only lead the user to the correct shelve but not the correct shelf level or actual item itself. Right now the problem is countered by voice guidance to the user as to where exactly the desired item is. However, to install a robotic arm on the RoboCart will have safety considerations for the user, as it might injure the user accidentally.

Assistive said...

u0204951 Xu Lixia
I was wondering if there is anyway to identify the robo cart belongs to a certain particular person.If there is no identification,then what happens if the robocart is stolen by other people?The visually imparied person seriously would not be able to know when his robocart is stolen or being exchanged with a normal trolley.And when he/she knows it,it would be too late.Its not wise to put identicfication number into the robocart as the personal identification details would be divulged to others once it is stolen.Maybe can consider making the robocart to be able to be folded into a pocket size for portability and when shopping,it can be clinged on to the user by some device .But this could also pose a problem as the robocart by being clinged on to the user,it may hinder the movement of the user or even cause accidents.

Assistive said...
This comment has been removed by a blog administrator.
Assistive said...

u0204714 Chan Hongjiang

Lixia has brought out an interesting point about the RoboCart being stolen by others. However, since the RoboCart can only work in an environment that has been fitted with the necessary infrastructure, ie the RFID tags, the painted lines on the ground etc, it would be useless for anyone to use it elsewhere. Thus, it wouldn't make sense for anyone to "steal" it. Also, I believe the RoboCart will be marketed to the supermarket and not to the individuals. I wouldn't want to spend my money on a RoboCart that I can only use in a particular supermarket.

As for the other comment on making the RoboCart "foldable", I seriously do not think that is a feasible idea. The RoboCart has to be sufficient big and powerful to be able to move around while carrying a load of groccery. So to "fold" it and carry it in one's pocket doesn't seem like a good idea at this point.

Industry said...

Neeti Warrier

I agree with Hongjiang in that the RoboCart would have to be supplied by the supermarkets themselves as it needs to be familiar with the layout of the shopping floor and the goods. Though customization through the use of identification cards is a good idea..

The post mentions that the user is to convey instructions to the shopping cart via a keypad. I think it would be better to use speech recognition to receive instructions. A visually impaired person can be spared having to punch out instructions which i suppose would be using Braille?

Assistive said...

U0204714 Chan Hongjiang

In response to Neeti's post, I would like to clarify that speech recognition was one of the input methods considered in the design phase of the RoboCart, however, it was not chosen for implementation as the recognition is unrealiable at times. Also, a Braille input pad was found to be too costly. The solution is a combination of a numerical keypad with synthesized speech. The keypad has buttons from 0 to 9, much like a telephone, as it was found that most of the visually-handicapped will be comfortable with using such a keypad. The keypad will have a special notch on the '5' button for easy recognition by touch. Then, the instructions corresponding to each button will be read out to the user using synthesized speech so that the user can input the desired operation. All of these information can be obtained from the 2 links in the original post :)