Smart Manufacturing

I work for a manufacturing company and we build large electrical and mechanical machinery with application to industrial robotics using PLCs (not autonomous robots).
Our use case is to develop a model predictive control that is trained continuously on industrial sensors data and do multi-variable forecasting, with couple of AWS services. I have evaluated Jetson Nano and Xavier industrial for industrial robotics. However, the architecture involve quite bit of threading and couldn’t prove suitable.
Now, I am looking into NVIDIA IGX Orin and other NVIDIA devices for this use case in industrial robotics as one hardware for edge computing and streaming data from and to the machine. And, we want to further automate and smart our manufacturing processes (laser, bending, assembling and multi-location inventory management.
Any one has attempted (or something similar) to develop model predictive control and smart manufacturing using NVIDIA IGX Orin hardware or any other NVIDIA AI hardware, besides Jetson Nano?

There are many interesting projects sharing at Latest Jetson & Embedded Systems/Jetson Projects topics - NVIDIA Developer Forums, some of them might be related with your projects, hope that useful to you.

Thanks for the reply. I have already searched through the Jetson and Embedded Projects. There have been some project using PLC, but with Jetson Nano device and we have found this device not suitable for our use case. And other most of the projects are computer vision based.

That’s why I am evaluating NVIDIA IGX Orin and other NVIDIA devices. On paper, they seems suitable, but I need real world evidence if some one had any.

Hello @JuwairiahZ,

Your project seems very interesting.

A couple of questions, if you don’t mind:

  1. What are you major concerns ?
  2. What requirements are you looking to cover with the Jetson device?


Hello @proventusnova and thank you for the response.

Our major concern is that we don’t want too much of hardware and software threading to develop model predictive control for our machine. Other concern for solution is to be enterprise level, high edge training and inference power on edge, and able to work in harsh weather and rigid conditions.
Our requirements are: streaming data from our sensors to a database, training model predictive control, inference and streaming it back to sensors. We are looking for a possibility to achieve this through a NVIDIA AI hardware and software. Our ultimate goal is to have one programmed NVIDIA hardware with model predictive control, which we can connect to any grain dryer sensors and then it starts intelligently grain drying after training model and inference on edge.

If this gets successful, then we can utilize AI for quality control.


Hello @JuwairiahZ,

Thanks for reaching back out and sharing all those details.

Got it, yes, it seems like you have quite a bit task on your hands.
From our experience, the NVIDIA Jetson Orin board is extremely powerful and it seems like all the requirements for your project should be achievable with the device.

As per the durability, there are manufacturers that offer rugged version of the HW, which we believe would be suitable for your project.

Now, we understand at this point you might still be a bit hesitant to jump right into such a large project without having all the data you can have on hand.
For that reason, we would recommend doing an initial proof of concept with a scaled down version of the system you want to build.

Just to give you a bit of context, we are a tech startup that offers Software Development services. And our area of expertise is Embedded Software Development.

We believe we might have a great option for you to perform a very low cost and risk free proof of concept.

Would you be opposed to scheduling a call so we can go over the details of your project so we can determine if we would be a good match?


Thanks @proventusnova

Can you share your company website link and details?


Hello @JuwairiahZ,

Sounds great!

We will prepare a proposal for you and will send it asap.


Hello @proventusnova
It is better if you can share your company details for our basic evaluation. I don’t want you to invest time in creating proposal just in case right now.