Project

# Title Team Members TA Documents Sponsor
27 A DodgeBot System
 Chiming Ni
Feiyang Wu
Kai Wang
Nichen Tian
Yiyang Bao
 design_document1.pdf
final_paper1.pdf
final_paper2.pdf
presentation1.pptx
proposal1.pdf
video1.mp4
video3.mp4
 Timothy Lee
A DodgeBot System
Team Members:

Kai Wang (kaiwang6)
Nichen Tian (nichent2)
Feiyang Wu (fw14)
Chiming Ni (chiming2)
Yiyang Bao (yiyangb2)
Problem
Dodge ball is one sport that require people to throw ball to each other and the one who got hit loss the game, which means that this game can not be played alone. However, sometimes athletes may want to take trainings to enhance their skill without compainions.

Solution Overview
Our solution is to create a dodgebot that act as another player. The dodgebot has a shooter that shoot the dodge ball towards the human player who is captured by the robot camera and detected and tracked by on bot artificial intelligence. When the ball is thrown towards the robot, the movement of the incoming ball is also captured by the camera and the dodging system will take contorl of the robot movement to avoid the collision.

Solution Components
Dodge Ball Shooting System
3-DoF Gimbal Design

Pitch Axis: DJI GM6020 motor + linkage mechanism (30° range).

Yaw Axis: Unitree servo motor (360° continuous rotation).

Launch Mechanism: High-speed pneumatic cylinder + pusher plate.

Actuation

Compressed air (0.5MPa) or Friction wheel drive

Adjustable launch angle to achieve 2 m/s initial velocity and 1.2m launch height.

Human Pose Estimation, Tracking and Dodging System
Jetson Nano (Or other edge computing platform counterpart)

A camera to detect people

Deep neural network trained to estimate human pose and detect incoming balls

Tracking system that output motor joint angles to the shooting system

Decision system that decide the direction of movement to avoid ball collision

Criterion For Success
The dodge ball shooting system must be shoot with a appoximately 2m/s initial velocity and be able to hit a person.

The human pose estimation and tracking system should be able to estimate human pose with a camera and track movement of the person with higher than at least 60% accuracy. The system should also output the correct angles for the motors to execute.

The dodge system should be able to move the hit dector at 2 freedom degree to avoid hit from a non-proficient human player.

Low Cost Myoelectric Prosthetic Hand

Featured Project

According to the WHO, 80% of amputees are in developing nations, and less than 3% of that 80% have access to rehabilitative care. In a study by Heidi Witteveen, “the lack of sensory feedback was indicated as one of the major factors of prosthesis abandonment.” A low cost myoelectric prosthetic hand interfaced with a sensory substitution system returns functionality, increases the availability to amputees, and provides users with sensory feedback.

We will work with Aadeel Akhtar to develop a new iteration of his open source, low cost, myoelectric prosthetic hand. The current revision uses eight EMG channels, with sensors placed on the residual limb. A microcontroller communicates with an ADC, runs a classifier to determine the user’s type of grip, and controls motors in the hand achieving desired grips at predetermined velocities.

As requested by Aadeel, the socket and hand will operate independently using separate microcontrollers and interface with each other, providing modularity and customizability. The microcontroller in the socket will interface with the ADC and run the grip classifier, which will be expanded so finger velocities correspond to the amplitude of the user’s muscle activity. The hand microcontroller controls the motors and receives grip and velocity commands. Contact reflexes will be added via pressure sensors in fingertips, adjusting grip strength and velocity. The hand microcontroller will interface with existing sensory substitution systems using the pressure sensors. A PCB with a custom motor controller will fit inside the palm of the hand, and interface with the hand microcontroller.