Seminar: Building a Robot Soccer Team, Spring 2017
Robot soccer has been adopted as a test bed and progress marker for
Artificial Intelligence and Robotics research since 1996. The primary success of
soccer as a test bed is that the high-level evaluation of a successful team is
readily understandable to everyone: the team that scores more goals, and “plays
better soccer”, is likely the team with the most effective innovations.
The goal of this 3-credit seminar is to explore the different sub-problems of AI
and robotics as applied to robot soccer, to survey the state of the art, and to
implement the most promising approach from each sub-problem.
The expected outcomes of the class is a group understanding of the state of the
art in the sub-problems of robot soccer, and to have a codebase with
high-quality implementations of the state of the art, which will then be used
for the UMass RoboCup team at the 2017 RoboCup competition at
Nagoya, Japan . The league that we will be entering in is the Small Size
League (SSL). Rules, descriptions, and team description papers of teams from
past RoboCup competitions are available on the RoboCup SSL Wiki.
There are no formal prerequisites for the class. However, to succeed in the
class, participants must be intimately familiar and experienced in algorithms
and data structures especially as applied to AI, linear algebra, vector
calculus, and coordinate geometry. Experience in programming in C++ is a strong
We will meet once a week. Every class participant must:
Select one sub-problem to study
Read and understand the papers listed for the sub-problem
Present a subset of the papers to the class
Implement the most effective approach for the sub-problem, in high-quality
All code must strictly adhere to the Google C++ style
guide, and must be consistent with the conventions for the rest of the code
base. All code must be well-structured and readable to any C++ programmer.
Individual functions must all be unit-tested.
Meetings: Fridays, LGRC A308, Times TBD
joydeepb+seminar [at] cs [dot] umass [dot] edu