Robotics Reference
In-Depth Information
signers, including schoolchildren, are encouraged to compete alongside
professional and academic groups.
Table Tennis
During the mid-late 1980s Russ Andersson built a table-tennis system at
Bell Laboratories in New Jersey that used a bat with an elongated handle,
held by a small industrial robot arm. Andersson devised a system of cam-
eras to track the three-dimensional motion of the ball, feeding the data
from this vision system to a planning algorithm that needed to know
the expected flight path of the ball. The planning algorithm calculated
the desired trajectory for the table-tennis bat and then modified this tra-
jectory repeatedly as the ball got nearer and nearer to the bat, thereby
increasing the accuracy of the data relating to the motion of the ball.
Andersson characterized the performance level of his robot thus:
The style of play emphasized precision, not the open power game
typical of human play. I think it fair to say that it beat the peo-
ple who wandered by to see a robot table-tennis demonstration,
but that competent players who regularly played table-tennis could
probably learn to be competitive after moderate additional practice
on that table. [2]
Despite Andersson's relative success, interest in the activity petered out
somewhat during the early 1990s without any researchers being able to
demonstrate anything remotely near to championship level performance.
Pool (Snooker)
Intuitively one would expect the task of building an expert level Pool
5
player to be relatively easy, given that the shots are taken only when all
of the balls are completely motionless, but despite this there are several
factors relating to the physics of Pool and to the strategy of the game that
are far from trivial to engineer and program. In a perfect world in which
the table was completely flat, the friction between the table and a moving
ball was uniform and the cushions and other features of the table had
absolutely no physical imperfections, and if several other conditions also
held true, then robot Pool would not be difficult. The software would
simply examine every plausible shot, taken from every angle, with every
5
The problems in designing a robot Pool player apply almost equally to all of the various games
of the Pool/Billiards/Snooker group.
