May 2024

Wheeled Bipedal Robot

A low-cost, DIY-buildable wheeled bipedal robot with 6 actuated joints, designed, built, and coded solo in 8 months for under $1,750.

mechanical·electronics·firmware·robotics

Role: Solo designer, fabricator & firmware engineer
Timeline: 8 months (Princeton senior thesis)
Stack: ODrive S1, Teensy 4.1, MPU6050, SolidWorks
Result: 12.4 kg payload capacity, $1,750 total build cost

Project Overview

Full CAD assembly of the wheeled bipedal robot

The goal of this project was to create a low-cost, easily manufacturable wheeled bipedal robot platform for use in research applications.

The robot was designed, prototyped, manufactured, and coded by myself in an 8 month timeline.

All components were designed to use either off-the-shelf hardware or be manufactured using easily accessible lab equipment including FDM 3D printers, laser cutters, and basic hand/shop tools.

Objective

Wheeled bipedal robots combine the strengths of wheeled and legged locomotion, offering fast, efficient movement on smooth surfaces and adaptability on rough terrain. Despite their potential, this field remains understudied.

This project presents a cost-effective, easily manufacturable design for a wheeled bipedal robot, intended as a flexible starting point for research groups. The design prioritizes affordability and accessibility, requiring minimal equipment for construction.

By offering a foundational hardware and electronics framework, this project aims to advance research in wheeled bipedal robots and foster further development in this emerging field.

Comparison of wheeled, legged, and wheeled bipedal robots

Design Process

Concept sketches

Initial sketch of knee linkage (top) and foot mount (bottom)

System diagram of electronic and hardware components necessary for full assembly

Early CAD

Electronics validation

Testing wheel balance by commanding velocity setpoints based on gyroscope data

Testing chassis stabilization by commanding position setpoints based on gyroscope data

Prototyping and assembly

First full assembly of knee joint with belt drive

Using cardboard to test chassis dimensions

All parts necessary to assemble the robot

Final Design

Challenges and Solutions

Foot Mount Splitting

Original heat set insert location (tension)

Improved heat set insert location (compression)

Challenge

The foot mount was splitting along layer lines due to the tensile load exerted on the plastic from the M3 screws

Solution

Move the position of the heat set inserts to the outer side of the mount
By moving the heat set inserts, the plastic experiences a compressive load from the M3 screw + insert, thus making the layer lines far less susceptible to splitting and cracking

Foot Mount Clamp

V2: tube was retained using a 4 part clamp

V3: tube was retained using a 2 part clamp with through-bolts

Challenge

The carbon fiber tube was slipping out too easily from the foot mount

Solution

In order to increase the frictional force acting on the tube, I knew that I either increase the friction coefficient of the materials or the normal force as dictated by the formula:

F_{\text{frictional}} = \mu N

In the second iteration, I made 2 design changes to the mount:
1. Create an inner and outer mount that clamps around the tube through M3 bolts that provide a compressive force
2. Add additional inner clamps that could be easily printed in a different filament material such as TPU in order to increase the friction coefficient between the piece and the tube
  1. Additionally, these pieces were smaller so they could be printed more easily in a finer resolution to achieve a better tolerance fit with the tube.
In the final iteration, I removed the inner clamping pieces because they were unnecessarily large and the compressive force from the screws was enough to retain the tube in place

Laser Cutting Tolerance Issues

Acrylic plates cut using a Glowforge Pro

3D-printed bearing retainer fitted into the laser-cut knee plate

Challenge

Tolerance issues from laser cutting kerf

Solution

3D print a retainer for the knee shaft bearing to fit inside the laser cut portion
This piece was small enough to quickly iterate on the exact tolerance necessary to interface with the laser cut hole

Hip Joint Mount

Original hip joint mount in the slicer, showing the support material required

Challenge

In order to make the hip joint easy to assemble, the original design used a ‘mortise and tenon’ joint which ended up requiring extraneous support material and being overly bulky.

Redesigned actuator housing, side exploded view

Redesigned actuator housing, angled exploded view

Solution

Redesign the actuator housing to allow for easier integration of the hip joint
This significantly reduced the size of the mount, allowing for a slimmer side-to-side profile

Results & Technical Details

PID tuning on a smaller scale version of the robot

Max Rolling Speed: 1.56 m/s
Max Payload Capacity: 12.4 kg
Total Weight: 11 kg
Total Cost: $1750

The robot was successfully able to be remotely controlled to move each joint with the wheels. Unfortunately, due to time constraints, the balancing PID controller required additional tuning and was unable to independently balance on its own. Future plans are in place to finish the PID tuning to allow for robot balance.

Hardware Details

ODrive S1 Motor Controller
- 20V power input
- Onboard absolute magnetic encoder
- Position, velocity, or torque control
- Communicated through serial communication

Eaglepower 8308 Brushless DC Motor
- Total Mass: 336 g
- Diameter: 92 mm
- KV Rating: 90
- Idle Current (24 V): 0.6 A
- Max Continuous Current: 22 A

Teensy 4.1 Microcontroller
- ARM Cortex-M7 at 600 MHz
- 7936K Flash, 1024K RAM
- 55 digital input/output pins, 35 PWM output pins
- 18 analog input pins, 8 serial, 3 SPI, 3 I2C ports

MPU6050 Inertial Measurement Unit
- Integrated 3 axis gyroscope and 3 axis accelerometer
- I2C communication

DeWalt Tool Battery
- 20V output, 6AH
- Built-in battery management system
- Simple charging interface

← Back to all work