For this Capstone project, we have worked with GovTech’s Sensor and IoT (SIoT) department. GovTech is a leading the digital transformation in Singapore and manages many of the government’s digital services while the SIoT department designs solutions to many of these digital services and collaborates with social services in Singapore such as Cerebral Palsy Alliance Singapore (CPAS).

As Singapore strives to be a smart city through the Smart Nation initiatives and to become a more inclusive society for everyone, we want to ensure that people with disabilities do not fall through the cracks. With around 100,000 wheelchair users in Singapore, including elderly and those with disabilities, we want to help these people to become more confident and independent with the development of a smart robotic arm.

Design Needs Identification Through Engagement With Stakeholders

Exploration and Evaluation Of Possible Designs

System Analysis Using Advanced Softwares And Simulations To Make Design Decisions

Taking into account the feedback from target users and stakeholders, our robotic arm solution consists of a highly mobile system that can be easily detachable from any kind of electric wheelchairs. Weighing at only 2.3kg and cost a little more than $1500 to prototype, its unique foldable design provides the user a low profile structural addition to existing electric wheelchairs. This enables the user to utilise the arm in public with an ease of mind.

Robotic Arm Operation Flow

The main page of user interface (UI) consists of preset floor values on 4 big and colourful buttons. These preset values can be saved for repeating the same operations in the future for convenience. Furthermore, the buttons in the UI are big and distinct for ease of pressing and differentiating.

Our image recognition module utilized YOLO v3 backbones and is trained on a digit-level dataset. The F1 score of this model on the test set could achieve 80.1% with reference time 0.469s. Thus, it is able to provide precise guidance for real-time arm navigation.

Moter torque

A simple workaround to include 2 motor instead of 1 to increase the torque output by 2 times. This will allow the base joint to be far more stable during operation.

Generalizing ability of the Computer Vision

As a further development of the generalizing of the CV, to allow the system to be more adaptable to different lifts and locations. Allowing the users to upload photos of their area’s lift panels to perform customized training. This would make our product more flexible to adapt to different lifestyles.

Open source initiative

Putting the project online with expanding the functionality, manufacturability and performance of the robotic arm. Increasing functionality in a way of picking up items. Increasing manufacturability to streamline the process of assembly. Increasing performance in terms of speed and accuracy in executing the given task.