By: Evan Gray '13 and James Conley '19
Advising Faculty: Gary Parker
All researchers face the challenge of obtaining sufficient test data to evaluate new systems for learning agent control. A number of factors, like cost and long development cycles, can dampen the progress of AI research when testing methods on actual robots. Methods for autonomous agent learning can be tested in interactive game environments as long as control of individual agents can be enabled and there is a reasonable means of testing the success of the agent controllers. Xpilot-AI is being developed as an effective test-bed for rapid development of intelligent control systems. Xpilot-AI provides an easy-to-use, inexpensive and flexible test environment.
Related Fields: Computer Science