Shooting for hardware cost less than $500

The arm is being designed from the ground up for this purpose. The arm will be kept light by mounting the motors all on the base, with long thin flexible shafts and bike brake cables actuating the joints. A raspberry pi zero w, controlling 10 or so steppers. Light, cheap, long reach take priority over accurate movement or rigid structure. Each 'bone' section of the arm will be extendible, allowing adaptive leverage to tackle both heavy objects like a pair of dirty pants close to center, as well as long reach for lighter objects like a feather duster. Grabbing will be primarily aligned visually, computing the relationship between the grabber and the object to grab.

2 cameras near the grabber like a freaking robot ostrich that serves you. Cheap stereo webcams, with simple structured light for easy rangefinding.

It will rely on its stereo camera vision for obstacle avoidance, looking at a spot before moving to occupy it. This will be processed on-robot for speed, while more complex recognition tasks will be offloaded to the cloud.

Skills brainstorm:

Potential apps

  • Is this trash? App
  • Close this door if it's left open
  • Any lightswitch is now connected
  • Go set the thermostat, I'm cold
  • See what the dog wants app
  • Stocking shelves
  • Shopping
  • Pick up dirty clothes
  • Training mode, repeat this motion
  • Action editor with timeline like audacity, add particular spots to aim for visually
  • Unload the dishwasher
  • Video game, it digs through fridge on its own, asking you if each thing is good

Types of apps

  • Go do something
  • Bring me something
  • Send me pics to make decisions
  • Everything is a smart thing

Arm design brainstorm:

  • use <$15 dremel flexible shafts to keep swung weight low on arm
  • use bike brake cables wherever possible, even cheaper than flex shaft
  • Use all the same stepper motors to reduce BOM
  • Extension/retraction arm for adaptable leverage