Rafiki Update

@Richard,

My source is RobotShop

Hokuyo URG-04LX-UG01 $1140 RPLIDAR $400

The one you mention can be sourced from ebay, Neato XV-11 Lidar replacement part less than $100.

If David is building a product, i didn't expect the Lidar to be a replacement part, what happens if the manufacturer stops the production for that old model ?

It seems too good to be true $100 compared to $1000

following my curiosity:

searching answers.ros.org for XV-11 a lot of people have issues with the performance (mapping, slam).

I'm only familiar with ROS & Kinect, did you tried the XV11 with ROS ? Do you recommend the purchase ?

After reading a few posts i think i m better served with the Kinect.

Lidar output, is a horizontal slice (array of distance/angle), not a point cloud, so if is located in the neck the area will be a horizontal slice around the neck, for slam asfaik you need to have transformations between the lidar and the base, so making the lidar movable (Pan,Tilt) is not so simple...at least for me

I need to read more about that, i'm only familiar with fixed sensors

The ASUS Xtion is better than the kinect... smaller, lighter and unlike the kinect it is powered via usb. No need for a separate power source...

I can only say that the final design is still in the works. This is a prototype to try to see what the best options are. I have 4 of the neato sensors and all have conducted the tests well so far, but that isn't to say that it will be in the final design. There is a lot of room in Rafiki to do a lot of things including adding a different type of sensor for SLAM. The lidar is a first attempt while also doing a lot of other development in many different areas. If it doesn't work out, the neck is an easy reprint. It is a lot better than having to reprint the body. Also, I might leave the lidar as a sensor for detecting objects that are higher and not for slam.

Quantity reduces the cost per item drastically, but building initially with expensive components isn't necessary. Proving the concept is more important initially. Cutting a hole in the body to fit a different sensor array might be in my future, who knows at this point. The design will certainly change over the next few months, but we are trying to keep that to a minimum, but are realistic in that reguard.

I do think that the lidar is the weaker of the two sensors for SLAM and might go with the other sensor in time, but only time will tell for sure.

@Richard

Yes, unfortunately i never found one in second hand with a nice price. I used a step up/step down with my 12v battery.

@David

i agree for a POC results are more important. I think the momentum is great for robotics, Intel Realsense sensor, ORBEC sensor and there is a stereo camera under 500, the DIY sensors like Lidar Lite

you can pick your poison

the main driver is a good motherboard/cpu.

Or ... having subsystems that handle a single thing, and report back only important information to the main robot computer.

This allows you to have a much lighter computer running specific systems like SLAM. The advantage to the lidar is that the information stream is lighter and more easially processed. If you only care about say something like a 60 degree field of view, it's really light weight and can be an even smaller computer. Granddaughter just woke up and opa is on duty... looks like I'm away for a while.

If you want to do SLAM you need a 3D Sensor or a Lidar:

1. Most 3D sensors need an USB3 port and heavy processing e.g. iCore (NUC), the exceptions are Kinect v1 (deprecated) and Asus Xtion Live (soon to be deprecated)

2. Lidar you can have a small systems (arduino, teensy, RPI), but the big question is: can a $100 Lidar perform like a $1000 Lidar ?

The Buddy Robot promises to have SLAM and other functionalities for $700, the main CPU will be an android tablet. I'm curious to know what kind of sensor they will use taking in consideration the robot cost.

Oh, believe me, I know what buddy is doing and what they arent. And I totally understand slam and what is needed to make it work.

I have been hesitant to say much but there are a lot of differences between buddy and rafiki. The largest of the is the funding that they have vs what rafiki has. Believe me when I say that there is no comparison. Here are some other differences...

Rafiki can have as many other sensors that you want to add to it. It is extremely expandable.

To use these sensors you don't need to code anything if it works with the EZ-B.

All of my robots will be programmible. There aren't seperate versions.

Rafiki will run for a much longer period of time between charges.

Rafiki will be a whole house robot platform that is expandable. Try telling buddy to come to you when you are at the other end of the house.

The robot is just one component of the entire system and isn't even needed to use most of the functions of Rafiki.

If someone wants to expand the slam capabilities of Rafiki, they can.

The access to external data is limited with buddy.

Buddy is about 1/2 the size of Rafiki.

Buddy runs an android OS. Rafiki can run really any os that supports serial ports. To use ARC for customizations, you can use windows.

If you want to run ROS, or anything else, you can with Rafiki.

If you want a completed robot we will offer that, but when you want to change it up or customize it, it is easy to do.

When parts fail on Rafiki 5 years down the road, you will be able to find them or replacement parts that will work. There is nothing that is a custom part.

There is a lot more that I can go on about. Buddy is a great robot, don't get me wrong. It is not that I think it is lacking in any way. I would buy one if I didn't have someone come to me and offer a very considerable amount of money for me to build Rafiki.

The 675k that they made from their funding effort won't be enough to design and build the molds, design the production boards and pay for the marketing engine. They will need more funds and will probably get it. Unfortunately I think they are short sighted and their robot will be seen as a toy by most who buy it. It won't be easy to upgrade and won't be able to easially grow in its abilities not only from the company selling it or by user customizations.

David,

Thanks for sharing your vision/ideas.

Regarding Buddy i think there is a lot of BS, i don't think it will do all the stuff presented at least the way presented in the marketing video.

I don't think is a consumer product, is not like open the box and the bot starts doing all the stuff, slam mapping, object recognition out of the box.

In my humble opinion Amazon Echo is the closest implementation of a consumer social robot, off course you have siri, cortana, google now on your phone but most people will see a phone.

the AI is already here in our hands.

for example NAO robot costs 8K, 700K for a social robot seems affordable, for example Alan only the (art work/plastic/silicone no electronics) costs the same price.

I think the key is always the price.

Do you have an idea what it will be the entry price/range for your Robot ?