System performance

Overview

Our system successfully met most of the requirements set out for us.

We succeeded in all design requirements and performance specifications save for the following:

• Robot must disinfect the injection region with a disinfecting wipe

• Robot must withdraw blood or inject the medication within the syringe

We also assess the following specifications as partially completed:

• Robots must be able to visualize the veins accurately, the acceptable error is +/-0.5 mm

Our success rate in settling on a vein site was 80%. We could locate veins in the imagery but there was noise in the segmentation and the injection location did not result in venipuncture (off by ~3mm from vein site).

Mechanical Subsystem

Our mechanical subsystem performed well on all requirements. 

Our control board succeeded in:

• Allowing independent control of all axis of motion

• Communicating with the stepper motors

• Communicating with the solenoids

Our stepper motors allowed:

• Accurate positioning along the clamp track and along the y-axis rail

• Traversing the clamping track

Our pneumatics allowed:

• Appropriate force to clamp the arm in place

• Appropriate force to secure the skin

• Appropriate force to puncture the skin and vein

Electronics Subsystem

Our electronics subsystem performed well on all requirements. 

We succeeded in:

• Providing sufficient lighting for vein segmentation

• Powering all actuators

• Powering our compute system

• Powering our control board

Perception Subsystem

Our perception subsystem performed well on most requirements. We were successful in detecting veins, however, it was prone to failing under different lighting conditions. False positives were also common in practice.

Neural Net Performance (given in Intersection over Union score):

38.73% Training IoU, 38.42% Validation IoU, 24% Test IoU

You may refer to our final report under "Documents" for qualitative results and analysis.