Teaching

Module Learning Objectives

By the end of the legged mobility module, students will be able to:

• model dynamics and control of legged systems using fundamental gait models;

• describe actuation strategies for legged robots;

• describe the core ideas behind state-of-the-art control approaches for humanoid robots, exoskeletons, and prostheses; and

• comprehend and summarize the main arguments of technical articles concerned with legged mobility.

Module Schedule

The module is organized around four topics: fundamental models, actuation, control of humanoids, and control of exoskeletons and prostheses.

Lecture 1 and 2: Introduction & Fundamental Models of Legged Systems

• Challenges and opportunities for legged robots 

• Key observations about legged systems in nature

• Standing dynamics and control (inverted pendulum model, balance strategies, capture point)

• Walking dynamics and control (capture point, landing impacts, speed limit, gait transitions) 

• Running dynamics and control (spring mass model, Poincare stability analysis, gait control)  

• Unified models for walking and running.

Lecture 3: Actuation in Legged Robots

• Overview of actuation in legged robots

• Electric motor units (dynamics, torque-speed relation, current control)

• Reflected inertia

• Series elastic actuation and variable impedance actuators

Lectures 4 and 5: Control of Humanoids

• Control using theory of kinematic chains (zero moment point, ZMP as stability measure,  walking patter generation with ZMP stability measure).

• Control combining basic functionalities (Raibert hopper, attitude, velocity and thrust control; swing leg mirror law, motor tapes)

• Optimization-based control (general controller hierarchy, MPC of center of mass behavior, instantaneous robot dynamics optimization with quadratic program, simulation example).

• Learning-based control (overview including general pipeline for reinforcement learning of locomotion control)

Lecture 6: Control of Exoskeletons and Prostheses

• Direct assistance (load carrying, effort amplification, effort prediction with deep neural networks)

• Mimicking normal joint behavior (motion replay, assist as needed, impedance and phase-based control)

• Virtual neuromuscular control (human motor control of locomotion, muscle-reflex based feedback control)

• Human intent integration (neural interfacing, indirect interfacing with neural network predictors)

Module Learning Resources

The module content primarily draws from research papers. Recommended papers and book sections are posted together with lecture notes.

Module Assessment

Students' comprehension of fundamental concepts and practical application skills are assessed through one assignment. The assignment is structured in several parts that align with the module's topic areas (fundamentals, actuation, control) and require the students to perform hand analysis and MATLAB simulation of basic gait models, actuation, and control methods for legged systems.