Enhancing actuator capabilities in human-centred robotic systems: New paradigms in gear train design and energy efficiency

Research output: ThesisPhD Thesis

Abstract

As human-centred devices gain traction and an increasing number of companies
aim to bring these innovative products to market, we are witnessing a pivotal
moment in the evolution toward true humanoids, advanced prostheses, and other
assistive technologies for both industrial and rehabilitation purposes. This surge
is largely driven by the advancements in Artificial Intelligence (AI), which has
revolutionised how these devices interact with their environment, surpassing what
was once thought to be science fiction. However, this technological advancement
has also highlighted a critical bottleneck: the demanding performance require-
ments of the actuators that power the joints of these devices. While electric
motors have a power-to-mass ratio that significantly exceeds that of human
muscles, their torque-to-mass ratio presents a substantial challenge, leading
to compromises in movement speeds, assistance or operating time. Typically,
these motors are linked to transmissions that convert excess speed into torque,
aiming to enhance the electric drivetrains’ torque-to-mass ratio. Yet, the currently
available industrial mechanisms do not allow for the required conversion amount
without becoming bulky, inefficient, or adding weight. This contrast underscores
a vital area of research and development, as the search for drivetrains that can
match this specific torque of human muscle becomes increasingly essential in
realising truly fully functional, responsive, and efficient humanoid and assistive
devices.
This thesis addresses this challenge by emphasising two key aspects: increas-
ing the torque-to-weight ratio, also referred to as specific torque, and reducing
electrical energy consumption in human-centred actuators. One approach is using
high gear ratio transmissions to optimise specific torque. After comprehensively
reviewing existing technologies, a novel Compound Planetary Gear Train
(C-PGT) based on the Wolfrom topology is introduced. This innovative design
aims to boost the gear ratio of conventional planetary gear trains. While the
Wolfrom topology offers promising features such as compactness, independence
of gear ratio from load capability, and reduced weight, the output-to-input energy
ratio –i.e. energy efficiency– drops drastically at high gear ratios. To overcome
this limitation, a novel design modification is proposed that introduces additional
design flexibility, significantly improving gearbox efficiency and torque capacity.
This work establishes a comprehensive design framework for this Compound
Planetary Gear Train, including fitting conditions, an efficiency model using
the rolling power concept and internal load analysis. It is shown that this new
topology results in an energy efficiency of more than 75% for a gear ratio of
143:1, with only plastic gears.
Another approach explored in this research for enhancing energy efficiency
involves incorporating elastic components. Series Elastic Actuation (SEA) and
Parallel Elastic Actuation (PEA) are recognised for their potential to reduce
energy consumption and enhance specific torque in robotic applications. SEAs
reduce geartrain stiffness, thereby increasing shock tolerance and enabling precise
force control. They also act as mechanical energy buffers, storing and recovering
mechanical energy during repetitive tasks. PEAs reduce energy consumption by
sharing the load torque, resulting in decreased motor torque requirements. This
study introduces an actuator with an elastic element connected to two of the three
shafts of a standard planetary gear train combined with dual motor actuation,
named the Spring-Embedded Planetary Dual-Motor Actuator (SEP-DMA). The
dynamical model has been developed for all possible configurations, serving as
a tool to characterise and study the proposed concept. This model facilitates a
detailed understanding of the optimal conditions under which each configuration
performs best. Using an offline optimal control algorithm, the power distribution
was calculated, and results show that the energy losses are reduced up to 40%
compared to a stiff actuator.
Additionally, robotic applications often require actuators to function under
two distinct conditions: high force and low speed, as observed in ankle prostheses
during propulsion, and low force with high speed, as seen during the swing
phase. Using a single actuator with a fixed transmission ratio leads unavoidably
to over-dimensioned actuators and excessively heavy drivetrains. Leveraging
on the accumulated expertise in high gear ratio transmissions and Dual Motor
Actuators (DMAs), this study explores the extra kinematic and static Degrees
Of Freedom (DOF) offered by C-PGT to reduce the weight and increase energy
efficiency. The theoretical framework developed offers insights into the DMA’s
vii
operational efficiency and mechanical design. An online control architecture
is proposed for the distinct operating conditions, and a proof-of-concept has
been developed and constructed. Experimental results validate the theoretical
models, showcasing the prototype’s power distribution and efficiency across
multiple operating conditions. The proof-of-concept demonstrates an 18% weight
reduction compared to a single motor equivalent capable of delivering a similar
output torque and speed and shows good torque tracking.
In conclusion, this dissertation presents several innovative approaches to en-
hance specific torque and decrease weight in actuators aimed at human-centred
robotics. Although the Dual Motor Actuator (DMA) concepts demonstrate con-
siderable potential, further research is essential to enable their transition to in-
dustrial applications. Meanwhile, the inventive C-PGT concept has resulted in a
successful patent family, currently on its path to valorisation via a spin-off com-
pany named Ailos.
Original languageEnglish
Awarding Institution
  • Vrije Universiteit Brussel
Supervisors/Advisors
  • Verstraten, Tom, Supervisor
  • Lefeber, Dirk, Co-Supervisor
Award date23 May 2024
Publication statusPublished - 2024

Fingerprint

Dive into the research topics of 'Enhancing actuator capabilities in human-centred robotic systems: New paradigms in gear train design and energy efficiency'. Together they form a unique fingerprint.

Cite this