Optimizing Robot Design: Heuristics for Efficient Performance

[Erik Bethke]

Hello all,

I am designing a lay-person-facing simulation that includes robots performing a broad array of tasks.

I do not want to create magical can-do-anything robots that collapse all work to be performed to a simple abstraction.

Rather, I would like to have relatively dedicated and optimized robots for particular jobs:
Drilling
Digging
Hauling
Smelting
3D printing [of material M, of size x, y, z and speed s]
etc...

I have a background in aerospace engineering and so I can comfortably calculate the theoretical watt-hours to accomplish any of these tasks. What I do not have a lot of experience with is estimating the losses in a robotic system.

My general plan of attack is the following:
1) Identify the job that the robot is expected to perform
2) Break that job down into smaller sub-jobs (e.g. run the manipulator arm for time t, move mass m distance d, etc)
3) Try to identify losses - e.g. wheels in soft sand, atmospheric drag
4) Sum all of these smaller jobs and identifiable losses
5) Have a standard CPU, display & communications black box and make it equivalent to the electronics in a Tesla
6) Now I have some sort of estimate of the total watt-hours to perform the job & controls
7) Then multiply by 2? 3? By 1.25? What is the rule of thumb here?

What am I forgetting to think about?

Are there better rules of thumb rather than a crude blanket across all jobs?

There must be different rules of thumb for sub-tasks such as:
1) locomotion on hard surfaces vs soft surfaces
2) gripping tasks
3) digging & drilling
4) 3d printing
etc...

Another way to look at it perhaps is that when folks go about designing a robot to do a task, how do they size the motors and batteries on paper?

Cheers and thank you,
-Erik

 

[malemdk]

Instead of thumb rule it is better to design from the scratch and add for the losses

 

[Erik Bethke]

Oh, I should have added, I am not designing one or two robots.

But will be on the order of 200-1000 specific robots.

It is a software simulation to help solve a larger goal. The errors on the robots power needs if they are within a factor of 2 or so would be fine. I do not want to be off by an order of magnitude.

 

[berkeman]

Another way to look at it perhaps is that when folks go about designing a robot to do a task, how do they size the motors and batteries on paper?

What am I forgetting to think about?

I don't see acceleration numbers of your loads taken into consideration in your post so far. That is part of sizing the motors (both from a power perspective and speed perspective). It takes a lot more power to do things fast compared to slowly...

 

[Erik Bethke]

Acceleration! Thank you - that is a big one.

What do you think about building essentially a spreadsheet of sub-jobs each with their own costing. Then to create a robot I composite the sub tasks...

Crudely:
1) Haul mass for a distance of d meters on level, hard surface, in time t
2) Haul mass for a distance of d meters on inclined, soft sand in time t
3) Grind a mass of substance with a hardness of h, in time t
...
...
?

Robot XYZ does 10x of 1, 3x of 2, and 1x of 3...

 

[malemdk]

you have to take into consideration the mass moment of inertia of the links and the servo gear motor rotor inertia ,which rotates the corresponding links
to get fine tuned systems , otherwise the commands and executions will be differ greatly