AI in Robotics April 23, 2008 Michael Buynak

Layout Introduction to robotics Ethics Historical (lots of videos)  Past  Present  Future

About robotics 'Robot' – introduced by Karl Capek in R.U.R. (1921)  We’d call them androids  means ‘forced labor’  Robots conquer the world Electromechanical system that seems like it knows what it’s doing  Assembly lines  Sweeping floors  Driving Needs to have some sort of intelligence

Purpose? Increase the quality of life Perform tasks that are to  Dangerous  Repetitive  Precise  Hard to reach  Boring My philosophy  If it’s too much work for you, make a computer do it.

Ethics Humans rely on machines  Could they take over? The Matrix Isaac Asimov & Bill Joy  Hesitant about really smart robots Three Laws + 1  0) A robot may not harm humanity, or, by inaction, allow humanity to come to harm.  1) A robot may not injure a human being or, through inaction, allow a human being to come to harm.  2) A robot must obey orders given to it by human beings, except where such orders would conflict with the First Law.  3) A robot must protect its own existence as long as such protection does not conflict with the First or Second Law. Why the future doesn’t need us

Old robot arm Unimate  Robotic arm 2 tons 3 DOF  Assembly Line  Hard coded

Progression Denso  6 DOF  Still hardcoded though c

Roomba Somewhat intelligent  Follow walls  Avoid obstacles Including stairs and IR beams  Finds it’s way back to the charging station No real model of the world at large

Markov Processes Markov Property  You can predict the future and know the past from only the current Sensor data is infallible  Fully Observable  Markov Decision Processes (MDP) Data is deterministic Sensor data has inherent flaws  Partially Observable  POMDP Data is stochastic

Error Deterministic  Error may accrue geometrically Probabilistic model

Bayes Filter Take data Move around Take more data Make a model Use it and update it

DARPA Grand Challenge Traverse 150 miles of the Mojave Desert  $1 million to the winner 2004  Best distance was 7 miles 2005  Completed in 6:54 by Stanford University  Carnegie Mellon at 7:05 and 7:14

DARPA Urban Challenge Navigate an urban arena  Obey all traffic laws  Merge with other traffic  6 hours to complete 60 mile course 36 semi-finalists  6 teams finished  4 teams finished under 6 hours  Carnegie Mellon won at 4:10:20  Stanford at 4:29:28

Walking robots Boston Dynamics  Big Dog  Cool legs  Smart stability, walking 

Star Wars Deka’s ‘Luke’ arm  Deka lead by Dean Kamen  Intelligent user interface  Most advanced robot arm ever?

Where it’s going Interaction with humans  In your home  At work Could become revolting before it’s usable  ‘Uncanny Valley’

SMU Robotics Autonomous Underwater Vehicle (AUV)  Competition in August in San Diego 11 th place last year  Best new entry  We beat MIT, UTD, CU Boulder, Duke... If you’re interested, we’ve got cool toys

Fun Stuff How to survive a robot uprising  Microsoft Robotics Studio  us/robotics/default.aspx us/robotics/default.aspx SMU robotics  /usv-home-page /usv-home-page

Questions or Comments?

References Why the future doesn’t need us  Unimate  Carnegie Mellon Peg Bowling  Roomba  Swarm robotics  DARPA Grand Challenge winning moment  Boston Dynamics ‘Big Dog’  Dean Kamen’s Luke Arm 