Introduction to Robotics, Marc Toussaint—February 4, 2014
Newton-Euler, joint space control, reference trajectory following, optimal operational space control
• Planning & optimization
goal: planning around obstacles, optimizing trajectories
Path finding vs. trajectory optimization, local vs. global, Dijkstra, Probabilistic Roadmaps, Rapidly
Exploring Random Trees, differential constraints, metrics; trajectory optimization, general cost
function, task variables, transition costs, gradient methods, 2nd order methods, Dynamic Pro-
Tunicates digest their brain once they settled!
• Motion was the driving force to develop intelligence
– motion needs control & decision making ↔ fast information
– motion needs anticipation & planning
– motion needs perception
• Control Theory
theory on designing optimal controllers
Topics in control theory, optimal control, HJB equation, infinite horizon case, Linear-Quadratic
– motion needs spatial representations
optimal control, Riccati equations (differential, algebraic, discrete-time), controllability, stability,
• Manipulation requires to acknowledge the structure (geometry,
eigenvalue analysis, Lyapunov function
physics, objects) of the real world. Classical AI does not
• Mobile robots
goal: localize and map yourself
Robotics as intelligence research
• Machine Learning and AI are computational disciplines, which
had great success with statistical modelling, analysis of data
State estimation, Bayes filter, odometry, particle filter, Kalman filter, Bayes smoothing, SLAM,
joint Bayes filter, EKF SLAM, particle SLAM, graph-based SLAM
sets, symbolic reasoning. But they have not solved autonomous
learning, acting & reasoning in real worlds.
• Neurosciences and psychology are descriptive sciences, either
on the biological or cognitive level, e.g. with geat sucesses to
describe and cure certain deceases. But they are not sufficient
• Is this a practical or theoretical course?
to create intelligent systems.
• Robotics is the only synthetic discipline to understand intelli-
“There is nothing more practical than a good theory.”
gent behavior in natural worlds. Robotic tells us what the actual
problems are when trying to organize behavior in natural worlds.
• Essentially, the whole course is about
reducing real-world problems to mathematical problems
that can be solved efficiently
There is no reference book for this lecture. But a basic wellknown standard text book is:
• Kinematics & Dynamics
goal: orchestrate joint movements for desired movement in task
robotics: mechanics and control.
Addison-Wesley New York, 1989.
(3rd edition 2006)
Kinematic map, Jacobian, optimality principle of inverse kinematics, singularities, configura-
tion/operational/null space, multiple simultaneous tasks, special task variables, trajectory interpolation, motion profiles; 1D point mass, damping & oscillation, PID, general dynamic systems,