Unit 12 · lesson
Select an End Effector from Evidence
Robotics videos make complicated grippers look automatically better.
They are not.
A mechanism with one moving part that reliably solves the task can be superior to a six-actuator hand.
Criteria first
Suppose the robot must collect objects that vary slightly in position and size.
Possible criteria:
- capture width;
- object damage risk;
- acquisition time;
- holding reliability;
- actuator count;
- mass;
- energy use;
- sensing requirement;
- ease of recovery after a bad pickup.
Now compare designs.
Example decision matrix
| Criterion | Weight | Roller intake | Pinch claw | Scoop |
|---|---|---|---|---|
| misalignment tolerance | 5 | 5 | 2 | 4 |
| hold security | 4 | 3 | 5 | 2 |
| simplicity | 3 | 3 | 4 | 5 |
| low mass | 2 | 3 | 4 | 5 |
Weighted matrices do not remove judgment. They expose it.
If you give a criterion a weight of 5, be ready to explain why it matters that much.
Recovery is part of the mechanism
Ask what happens when acquisition fails.
Can the robot:
- back away and retry;
- open wider;
- reverse rollers;
- detect the failed grasp;
- release a jam safely?
A mechanism without a recovery path can turn one small failure into the end of the mission.
Final selection
Create your own decision matrix with at least three mechanisms and five criteria.
Then write:
- selected design;
- strongest reason;
- most serious weakness;
- sensor or human evidence needed to confirm successful acquisition;
- recovery behavior.
Add the result to your dossier as a subsystem decision record.
Use a decision matrix, then challenge it
Suppose a robot must pick up three object types: a cardboard box, a plastic cylinder, and a soft bag.
Candidate end effectors:
- parallel-jaw gripper;
- suction cup;
- flexible scoop.
Score each from 1 to 5:
| Criterion | Weight | Gripper | Suction | Scoop |
|---|---|---|---|---|
| handles all shapes | 4 | 3 | 2 | 4 |
| gentle contact | 3 | 3 | 5 | 4 |
| simple sensing | 2 | 4 | 3 | 3 |
| low power demand | 2 | 4 | 2 | 5 |
| easy release | 2 | 4 | 5 | 4 |
A weighted score can organize reasoning, but it is not proof. A bad criterion list produces a confidently bad choice.
Prototype the uncertain assumption
If suction scores well but the bag surface may leak, the highest-value test is not another spreadsheet. It is a bounded test of seal reliability across representative surfaces.
Good design decisions move between models and evidence:
requirements → comparison → uncertain assumption → test → revised decisionRecord which assumption your next test is meant to reduce. That keeps prototyping from becoming random building.