Advanced topics are not in CS Academy, but are required
and can appear on any future quiz or exam.
However, even if these topics appear on a future assessment,
we would not expect you to have an understanding beyond what is
covered in these notes.
Lecture Notes:
Sense-Think-Act paradigm
Sense any new information from sensors (environment, robot state, external commands). Loosely analogous to updating the model in MVC.
Thinks, draws conclusions or estimates from the model, and plans an action, Loosely analogous to the controller in MVC.
Acts by moving motors, sending messages, or controlling other outputs. Loosely analogous to the view in MVC.
Force control vs Position control
Industrial robots often use position control for extremely fast and repetitive motions in predictable environments.
Older walking robots use position control to keep their center of mass or zero-moment-point above their support polygon at all times. These robots are stable but slow, and often appear stiff and toy-like.
Force control allows robots to operate more like humans, who can react to touch and can recover from unexpected collisions
Series-elastic actuators and compliant joints are better suited for precise force control than traditional geared motors.
Mike's Masters thesis on compact series-elastic actuators
(link).
More modern robots often use very high-torque direct drive motors with less gearing.
Limitations in robotics:
Battery and motor power density were barriers to building practical, agile humanoid robots prior to 2020
Fast, reliable computing hardware and control algorithms also limited the advancement of dynamic motion prior to the late 2010s
Concerns about human safety remain: Dynamic robots often fail in unexpected ways, and compliant joints do not necessarily reduce the danger of collisions with humans.
Modern barriers include reliable environment perception and interpretation, battery capacity for extended runtimes, and compact compute power for higher-level on-device planning with LLMs.