Raspberry Pi Forums

For more than 20 years I have wanted my robots to become independent - managing their own charging to allow 24/7 operation.

This year, with a GoPiGo3 robot, I substantially achieved this goal, and have learned more about the challenges facing a 24/7 robot. Specifically, in my case, the capacity of NiMH AA cells begins to diminish quickly around 140 to 150 cycles, 1200 to 1500 hours, or 2 months of 24/7 operation.

NOTES:

Batteries: Powerowl 2800mAH NiMH AA-cells

Charger: Tenergy 1025 Smart Charger set at 1A max rate

Typical Docked, Charge Time: 3 hrs

Playtime: Start at transition to trickle charging, end at 8.1v reported
(8.1+0.6 protection diode drop = 8.7 / 8 cells = 1.0875v/cell )

Load: Measured initial battery load of 318 mA at 9.6v

Configuration: Raspberry Pi 3B ave load 0.00 (uptime), GoPiGo3, 

TOF-DistanceSensor, USB powered speaker

Data Span: First 2 months 7-1-2019 to 9-6-2019

Life Point: After 156 docking cycles 1272 hrs operation in Carl

ANALYSIS:
- Minor capacity decrease until 145th cycle (10% or 0.07% per cycle)
- Steep reduction in capacity in next 10 cycles (10% or 1% per cycle)



.
- For Comparison: (candlepowerforums 2014 Eneloop at 500mA rate)
100 cycles: near flat capacity,
150 cycles: down to 90% capacity
175 cycles: “Knee to 0%” at 80% capacity


.
Several people have suggested Lithium Ion as a better power source, but the charging complexity of these cells is quite a hurdle to integrate into my two electrical contact docking system, and most of the Raspberry Pi oriented LiPO hats have GPIO header pin conflicts with the GoPiGo3.

I just noticed a possible non-conflicting hat called the PiJuice, which could work for powering the Pi (5v) but getting the 7.2v - 12v to drive the GoPiGo3 board, and the extra stack height, and isolation, and who knows what else, will remain challenging.

Designing real world solutions is so complicated…

A dc-dc buck converter will efficiently provide 5V to PI from a LiPo's 7.2 or 11V pack

They are cheap and small, since Lipo-s are more compact than NiMh you should find some room for them.

You need to closely monitor battery voltage to prevent overdischarge. There are also charging equalizers that are also small so you can pack them together with the batteries

https://www.amazon.com/s?k=buck+convert ... b_ss_i_1_6

battery charge/discharge protector: https://www.amazon.com/Anmbest-Charger- ... 257&sr=8-6

These are just examples you can find cheaper alternative examples
----------------------

PS It would be just as interesting to explain how the charger works mechanically and how the robot locates it and plugs in.

You should get a higher number of cycles with LiFePO4 cells.

There are some graphene LiPo's now that say they can charge more times.
HobbyKing's Turnigy Graphene are one brand saying this.

Super capacitors have made a breakthrough, potentially millions of cycles.
One reason Elon acquired Maxwell.

Energy density is the issue whether it is batteries, liquid fuels or nuclear power.
There are now guys playing around with combustion engine generators to keep drone batteries charged.
If you have a generator then supercaps are more useful.

The best combustion "generators" seems to be gas turbines.
Advances in those using super critical CO2 and ceramics means very small and high power density.
Can they be made small enough for mobile bots?

Super critical CO2 turbines are very useful can be also be used for stored solar thermal or molten salt reactors.
But that's the future, no need to worry about the AI/robot apocalypse till them

PS It would be just as interesting to explain how the charger works mechanically and how the robot locates it and plugs in.

That has not really changed since the days of the first mobile bots, a docking station and metal contacts.
Generally IR beacons these days.

We should be able to get Bluetooth 5.1 modules this year, with dual antenna base stations, cm scale positioning comes inside.
GPS RTK can do cm accuracy but only works outside.

Induction charging is used for charging phone 1-50W? etc they have issues with heating up metal objects.
Probably ok for small desktop roaming bots

A newer higher frequency AirFuel using 6.78MHz and GaN transistors is coming out of the labs.

Renewable? been some estimates a Pi need a 100W panel.
Some guys don't get enough for 24/7 from 30W
That's just for powering the Pi, not the motors etc.

It's a challenge.all right.

blimpyway wrote: ↑

Sat Sep 07, 2019 9:20 pm

A dc-dc buck converter ...
charging equalizers ...

PS It would be just as interesting to explain how the charger works mechanically and how the robot locates it and plugs in.

Yes, powering my bot would need some combination of converters, equalizers AND UPS function. The total requirements / constraints make designing a replacement an investigation of multiple alternatives.

Here is details, schematic, etc. on my existing system and a video of Carl getting on and off his dock

And this is the write up on Carl's current "Find the Dock" using OpenCV"

Gavinmc42 wrote: ↑

Sun Sep 08, 2019 1:12 am

You should get a higher number of cycles with LiFePO4 cells.
...
Some guys don't get enough for 24/7 from 30W
...
It's a challenge.all right.

The 8x NiMH AA cells fit in 4x4x1.25 inches and deliver 18-21 Watt-Hours (318mA idle * 9.6v nominal * 6 to 7 hours "fairly idle play time").

What allows 24/7 operation is that the combination of the NiMH batteries, the Tenergy smart charger, and the GoPiGo3 bot board allows simultaneous charging and powering the "fairly idle" bot while sitting on the charging dock.

The bot runs for 3 hours on the dock recharging.
When it detects trickle charging it gets off the dock for a maximum 6.6 hour playtime,
When it detects Vbatt approaching the knee (with a safety factor) it docks again.

(Right now I am not taking advantage of the bot being on 24/7 in any way. It is just that I don't like to shut "him" off. Mental case ...)

(Right now I am not taking advantage of the bot being on 24/7 in any way. It is just that I don't like to shut "him" off. Mental case ...)
Yes it is cruel to turn them off when they are trying to evolve AI genetic algorithms
Once you put a face with eyes on them they are individuals, like babies and puppies.

I would love to get my hands on some of Tesla's 21700 cells.
It is easier to get 18650 Li-Ion and LiFePO4 cells now, even my hardware store has them.
The old ones recycled from cordless drills/laptops are on their last legs.

I'm thinking of perhaps using 26650 cells.
Right now I'm trying to design a 100W wireless charger.
After that it is 1KW wireless one for AGV's.
So far it is looking like needing a class E transmitter.

Here is details, schematic, etc. on my existing system and a video of Carl getting on and off his dock

And this is the write up on Carl's current "Find the Dock" using OpenCV"

That's really cool, thank you