Hazard detection

Hi there,

I am trying to build a robot using rasp pi 5 that can patrol warehouse and detect the hazards. could you please suggest all the parts required including power option for rasp pi 5, sensors, chassis, etc

Thank you

1 Like

Hey @Ashim278021, welcome to the forums!

That sounds like a pretty neat project! I would start by outlining the main components of your project to try and get a good starting point, something like this:

Objectives
Outline the main point of the project in semi-specific terms, rather than: “I want to detect hazards”, specify; What kind of hazards? How accurate, and how many different kinds? How big should the device be?

Asking yourself these questions can very easily flesh out a project at the start, and get you going in the right direction.

Inputs
Refer to your objectives, what properties of the hazards could be exploited here? What kind of things are you sensing? This can lead you to the types of sensor options available.

Outputs
What is the end goal of the robot? Will it send a signal to shut down machinery? Will it activate an alarm? Will it construct a wall around the hazard? This step will also help structure the device and allow you to gain a deeper understanding of the final product.


Ideally, a large post explaining exactly what you want to achieve and how is the best place to start. If possible, include diagrams, images, comparisons, any additional material to allow others to gain a better idea of your vision.

A question I had in particular: Why have you chosen a Pi5 for this project? Not that there’s anything wrong with using a Pi5, but its possible the result could be achieved with something cheaper, and this way you can get thinking about your design decisions!

Looking forward to seeing a more expansive project brief, happy making!

3 Likes

Hi Zach,
I already have a rasp 5 with me thats why I want to use it and i just want to develop a robot that detects the hazard and sends a notification to your mobile.
For example: Hazards can be something thats blocking the walking path in warehouse or some suspicious individual or body present during night.

I want to create a simple detection robot not too complex please

1 Like

What you are asking is not impossible.

Where I am at Swinburne Uni, robotics/mechatronics is a complete 4 year Degree all of it’s own. That means, it’s complex with lots of maths. Machine Vision is a Semester all on it’s own.

Not trying to discourage you but what Zach is alluding to is really important. You need to define it in detail according to Zach’s outline (he’s probably been here already!).

Looking forward to your analysis.

Gerard

2 Likes

Hi Gerard,

Thank you so much for the information.
I am very new to this topics right now and I am not trying to build everything from the scratch. I just need to build a robot using some built in chassis, sensors, cameras and raspberry pi for detecting simple things.
I am not trying to use it in an actual ware house. Its only for my uni curriculum. So it does not need to be sophisticated.

Thank you

Kind regards

Ashim

1 Like


Here I have attached a similar project. I FORGOT TO MENTION AN AUTONOMOUS ROBOT. but I was a bit confused. how should I manage the power for rasp pi 5 and should I use Lidar or any other sensor so that it can navigate easily through the warehouse.

1 Like

Hi @Ashim278021,

That sounds like a decent place to start, If you are only looking to detect the presence of an object, an ultrasonic sensor sounds like a simple place to start for detecting when something is in front of you.

The Pi 5 has a power requirement of 5V and 3-5A depending on how many extra modules you have connected to it. For a moving robot like this, the motors are likely going to be your biggest power concern so I would look into that before you think about powering other components.

I would approach this by first planning out a rough size and weight for the robot. From there you can move onto moving it around, then finally onto the sensors and Pi setup.

With projects like these, it’s normally the case that you can solve the problem with a near-infinite amount of solutions. The more specific your requirements and goals are for the project, the shorter this list of possible solutions becomes.

Hope this helps!

2 Likes

Hi Samuel,

Thank you so much for the information.
Much appreciated.
I want to use this chassis (Yuewalker - Tracked Chassis | Buy in Australia | ROB0154 | DFRobot | Core Electronics) and use rasp pi 5 to control it with attachments like pi camera3 and a sensor(maybe Lidar or ultrasonic sensor). Could you please provide me some ideas for power management to control it and what motor controller should I go with it??

@Zach Could you please provide some insights??

Thank you so much

Kind regards

Ashim Thapa

2 Likes

Hey @Ashim278021,

That’s an awesome choice! The tank chassis looks really cool.

That chassis already comes with motors which is good. Working backwards from the motors you will need something that can run 2 DC motors at 12 volts. The Makerverse Motor Driver 2 Channel would be my first choice. This module also has a 5V out so you could use this to power the Pi assuming you provide it with enough current.

Alternatively, you could use a motor driver HAT with the Pi instead of a stand-alone module. Something like this 10A 6V-24V DC Raspberry Pi Motor Driver HAT may be a little cleaner of a solution.

The motors being provided with the chassis does remove some complication from this. I would be trying to find one part that does the job of taking the power from a battery to all of the components in this project. So far this looks like just the motors, the Pi and the parts that will get power from the Pi.

Hope this helps! :slight_smile:

2 Likes

Hi @Samuel

Thank yo so much.
Much Much appreciated.

1 Like

Hi Ashim,

On the topic of hazard detection, and localisation of the Robot - Gerard covers it well.

You might be able to get something that will get the job done, in a small time period but making it reliable will take a lot of effort.

If you are keen to venture down this deep rabbit hole (and be warned it is very very deep), I would consider using ROS, and a SLAM package for the lidar portion, and an off the shelf model for hazard detection.

Using lidar for ‘hazard’ detection is no small feat - I would consider this as one of the final steps in this project.

Liam

Hey @Samuel Do you think this (Raspberry Pi 5 18650 Battery UPS HAT (5.1V 5A)) will help me power the rasp pi 5 and the parts or
Do I need to look for more power??

I will consider that Liam.

Thank you so much for the suggestion.

You need to use RPLIDAR A1/A2 or YDLIDAR X4 for mapping and obstacle avoidance. For basic obstacle detection use HCSR04. For fire and heat detection use MLX90614 or AMG8833 thermal camera. You can use MQ-2, MQ-135 and SGP30 for gas, AQI and VOC detection.

3 Likes

Hi Ashim
What is the physical size of this warehouse.
Do you really need a robot wandering around.
Would a fixed system of detectors looking down the aisles etc simplify things as a moving source looks like it could be very complicated. Even a wall at the end of an aisle or any fixture is going to change relative position and unless some pretty smart software can be devised could be interpreted as a moving target.
Cheers Bob

2 Likes

Hey Bob

Thank you so much for the suggestion. I am not going to use it in the actual warehouse. It is just my uni curriculum where I need to build a robot that can detect things. I thought warehouse would be a suitable area to test it.

Kind regards

Ashim

Hi Ashim
That explains it. I really did think there would have been easier ways to detect people and things that were not supposed to be there.
Cheers Bob

Ashim,
I agree with Liam on this - it can be a very deep hole. I have been on a similar journey (just a winter hobby) for many years now and am on the ‘Mark 4’ version.
I could tell you the long story but that is almost a book now. Things that haven’t been mentioned yet but would have helped me if i had specified them first.

  • if its going to move around decide how fast this need to be. At about the end of Mark 2 I changed the gear boxes on wheels so it could move along much faster. It now hits walls easily because its wheels move faster than it can process the Ultrasonic return signal.
  • Try and buy sensors that use interrupts to get the attention of the ‘processor’ . You will go mad trying to ‘poll’ sensors.
  • Ultrasonic has a pretty wide field of view and Lidar more narrow. Neither will get over 3 or 4 meters. (If you get to change your assignment reduce from warehouse size to living room - dodging chairs and kids toys is still a great challenge)
  • Do you have to just dodge things or recognise what the obstacles are. Eg that is a crate that stays there vs its a person vs ‘that has moved…’.
  • Again if you can change the scope, I think it would be easier to imagine a Sentinel that stands in the middle and has a picture of what it expects to see, then can far more easily decide if the ‘picture’ is nor what it should be.
  • My toy had a lot of trouble telling the difference between ‘floor’ and ‘wall’. To a camera these are the same shape, and in my case colour. (More excuses to barrel into the wall). It was a couple of years ago but AI did not have any models in this domain (it seemed to concentrate of roads, fields etc). If you have to build your own data set for AI add a year or two to the project.
  • How long does it have to run for? Mobile power is its own monster. If it has to run for a whole ‘shift’ that spells trouble. Maybe you could base it on a robot vac - at least that knows how to go home for ‘feed’.

And just for context - my Mk4 is a split processing design with the Mk2 Arduino doing all movement a simple sensor processing. On top of this is a Jetson Nano doing the video work and associated AI image recognition.

Good luck
JC

1 Like

Hi John,

Thank you for the guides.
I really appreciate it.

Kind regards

Ashim Thapa