How WE RIDE Bicycles

Introduction

Human control of bicycle dynamics with experimental validation and implications for bicycle design

Researchers

Dale L. Peterson
Jason K. Moore
Ron Hess
Mont Hubbard

Sports Biomechanics Laboratory
Mechanical and Aerospace Engineering Department
University of California, Davis

Ability vs. Understanding

(Almost) everybody knows how to ride a bike, but (almost) no one knows how we ride a bike

Project goals

Understand what we do while riding a bike.
It doesn't study "How people learn to ride a bike"
but rather "What they are doing after they know how"

Motivation: Theoretical and Practical

Bicycles are an economical platform for understanding human control of dynamic systems.

Roughly half our energy use and carbon footprint is for transportation.
More than 50% of our trips are less than 2 miles, ideal for bicycles.
Bicycling and walking will eventually become a larger share of our transportation solution; sooner is better.
Knowing how we ride, and how bike design affects this, can help in bicycle design for outlier populations like elderly and children and promote more biking.

How can we do this?

What is a dynamic system?

A system that changes in time. It has:

State variables
Inputs
and often Control

Example dynamic system with feedback control

Balancing stick or broom balancer
Differential equations describe how the stick falls
Human senses position (and much more) and executes corrective action to move hand

Bicycle Human Feedback

On the bicycle, human sensory organs include:

visual system
inner ear otolith organs (vestibular)
tactile sensors in skin (proprioceptive)

and sense quantities like

rotation rates
limb position
speed and direction
error in position
etc.

Bicycle Human Feedback

Brain integrates all this information in real time to produce a single control signal that actuates muscles to turn handlebar. This is the process we study.

Our project is famous and award winning!

Washington Examiner

Our project is famous and award winning!

Senator Tom Coburn R-OK

Taxpayers may also question the value of many of the projects NSF actually chose to fund, such as: How to ride a bike...

Organization

Luke Peterson
Bicycle dynamics without rider
Ron Hess
Theory of human control
Jason Moore
Experimental identification of rider control

Bicycle Dynamics 101

What/Why/How are we modelling?

What: Motion – Dynamics
Why: Understand the influence of physical parameters on “stability” and “handling”.
How: Apply Newton’s Laws to obtain differential equations which describe how the bike moves.

Modelling Assumptions

Start simple!

Rigid parts (two wheels, frame and fork)!
Rider rigidly attached to the frame!
Frictionless joints!
No slip rolling!
No aerodynamic affects!

Even with these simplifications, we still have 25 parameters to keep track of.

What do we observe in real bicycles?

Is the motion stable?

Eigenvalues vs. Speed

Unstable at 0 km/h (0 mph)

Unstable 12.6 km/h (7.9 mph)

Stable at 17.6 km/h (11.0 mph)

Unstable at 22.7 km/h (14.2 mph)

Stability of upright motions

Counter-steer

Unconventional bicycle with stability

What is observed? Unconventional bicycle with stability

Turning Bicycles

Stability of turning motions employs same mathematical framework

Key Differences

Straight line riding is an equilibrium at any speed, perhaps stable in a range of speeds. No rider steering input need for stable speeds.
Turning bicycle: for constant v, turns of different R will require different lean, steer, and steer torque. Rider must apply constant steer torque.
More difficult to test experimentally.
Stability of turning motions is an active area of research.

Robotic Bicycle

Why?

Eliminate confounding factors of human rider.
Quantitative comparisons of model and experiment.
Test different feedback control techniques.

Robotic Bicycle Experimental Setup

Robotic Bicycle

System identification of upright motions and turning motions:

Apply precise inputs (motor torques)
Measure response (angles/rates/accelerations)

Test the validity of modelling assumptions:

Under what conditions is the pure rolling assumption reasonable?
Is rigid-body assumption reasonable?
Are frictionless joints appropriate?

Improve the model as needed.

Challenges

Real time data acquisition, control, logging, remote monitoring.
Detecting "crash" conditions and responding safely.

Theory of Human Control

Bicycle + Rider = Balance?

Feedback Model

Basic model will include:

Visual, proprioceptive and vestibular sensory feedback
Rudimentary neuromuscular model

Six Bicycles Modeled

How pilots rate aircraft “handling qualities”

Handling qualities and the helicopter pilot

Predicting a bicycle's handling qualities

V = 2.5m/s

V = 5m/s

V = 7.5m/s

Bicycle Control at Slow Speeds

Professor on a bicycle

Rider Model Results in Lane Change Maneuver

Roll angle with rider control.

Rider Model Results in Lane Change Maneuver

Steer angle with rider control.

Rider Model Results in Lane Change Maneuver with Hands Free

Rider model + bicycle in lane change with hands-free

Identifying the Rider + Bicycle System in Experiments

Instrumented bicycle

Identification results rider + bicycle roll-rate response to pulsive side forces

Experimental Identification of Rider Control

Start Experimenting: On the treadmill

Professor can now balance!

Bike Snobbery

And with that, I’d like to leave you with a quiz. As always, study the item, think, and click on your answer. If you’re right, you’ll see some sort of confirmation, and if you’re wrong you’ll see evidence of a sinister Dutch plot to create a legion of deadly supercommuters.

– Bike Snob NYC