[Clay Builds] found a bit of a gem at a recent auction, picking up a Nomadic Technologies N150 robot for just $100. It actually looks like something out of science fiction, with its cylindrical design, red bumpers, and many sensors. He decided to try and restore the research-grade robot to functionality with the aid of modern hardware.
Right away, it’s clear this was an expensive and serious bit of kit. It’s full of hardcore gears and motors for driving three rubber-tired wheels, each of which has a pivoting mount for steering the thing. Through his research, [Clay] was able to find some ancient websites documenting university work using the robots. His understanding is that the platform was designed for researchers experimenting with simultaneous localization and mapping (SLAM) algorithms, and other robotic navigation tasks.
[Clay] doesn’t just settle for a teardown, though. He’s been able to get the platform running again in one sense, using an Arduino to manually run the robot’s drive controls under the command of a gamepad. Without official software or resources, it’s perhaps unlikely he’ll be able to get the stock hardware to do much without completely rebraining it, so this method makes sense. In future he hopes to get the bumper sensors and sonar modules working too.
It’s a fair effort given [Clay] was working with no documentation and no supporting software. We’ve seen similar efforts for robotic arms before, too. Video after the break.
Looks like a fancy Heathkit Hero-1
That form factor robot was made by 3 companies around that time. The company near Boston was Denning Mobile Robotics and I’m familiar with them. There was a series of company-bought-out-by-other-company, but I *believe* DMR was the company that made the original trashcan design with the ultrasonic range sensors – essentially in the form of a Dalek.
The CPU chip on the motherboard is an MC68332, which was a high-end Morotola microcontroller chip, essentially the microcontroller version of the MC68000 series, and you can find datasheets for that online.
The chip with the label is a UV erasable memory chip. It was standard practice to erase the chip (UV through the window for a specified amount of time), program it with software, then put a label over the window and write the software version on the label. The label protects the chip from accidental erasure from ambient UV such as sunlight. These are easy to read, just pop them out of the socket and grab a cheap $25 EPROM reader from eBay.
The processor was almost certainly running a bespoke code programmed in C, with pieces written in assembly. It’s almost certainly not running linux or anything big, so you could probably download the data from the EPROM chips and disassemble it. 68000 code is not difficult to reverse engineer.
The code on the motherboard is almost certainly a simple serial control and data reading protocol, so that the actual processing could be done remotely on a a “big metal” computer that then sends commands to the device. Put an oscilloscope on the serial port TxD and measure the time between pulses, invert that number and divide by number of bits to get the baud rate. Connect a serial converter to the serial port and play around with it – typing “?” might give you a list of commands. Figuring out the serial protocol from reverse engineering the code should be straightforward – it’s probably single character codes dispached from an N-way branch (ie – select statement) inside a fairly small main loop.
The purpose of the company (Denning Mobile Robotics) was to find a use case for an autonomous robot. Microcontrollers were fairly new, and putting one on a mobile platform was a solution looking for a problem. DMR was trying to sell these to run around a company warehouse at night looking for intruders. IR beacons were installed at points on wall, and the robot would go to the beacon, turn, go to the next beacon, and so on… all the while looking for people.
One proposed application had a gun mounted on the robot. Not making that up.
Ultrasonic range finders were originally developed for Kodak cameras, and were adapted for robotic use, and have a ton of problems. Ultrasound goes in a straight line and bounces in the manner of a light beam, so hitting a surface at an angle the beam won’t reflect back to the sensor. Then, it can reflect off of the wall and also reflect off of a 2nd wall forming various ghost signals, and there’s error in the resolution of the reflected signal timing, so that you can’t simply plug the response times into a matrix and solve – you need the solution that minimizes the error on each measurement, and that’s a mathematically hard problem.
I never heard of DMR ever selling the devices to anyone for actual use. State of the art at the time was following beacons, because the unit couldn’t figure out its position using the range finders.
I don’t know that anyone uses any of the robots from Boston Dynamics either, and I think they have a model that will patrol a warehouse floor at night as well.
I’m pretty sure those ultrasonic transducers were developed for Polaroid, not Kodak …
Dalek? Oh, must extereeeeennninate! For real, I remember reading about it.
pretty good writeup and explanation of how to get one of these things up and running, talking with modern equipment over a serial port
There were the trashcan robots from the movie Malcom, they were radio controlled…
You probably already found this site:
https://drobot.sourceforge.net/faq.html
It lists some of the Nomadic types, including the N150.
I was in college in computer science in the 90’s. We had two Nomad robots used in the software engineering class. The entire class was split into maybe a dozen teams and had to share the same robots throughout. We signed up for slots to use them, sometimes in the middle of the night. Otherwise, we could only test code on a software simulator. Needless to say, real life didn’t quite match the simulator. Having an available set of charged batteries was a constant challenge. I can still hear the ticking of the sonar sensors. It wasn’t nearly as cool as I hoped. Fun times!