SBIR Phase II: An Interactive 3D Game of Evolutionary Robotics for STEM-based Education

这个SBIR第二期项目研究如何通过虚拟机器人计算机游戏中的基于游戏的学习来吸引学生参与STEM（科学技术工程和数学）教育。在游戏中，学生们设计、组装和训练虚拟行走机器人，与其他机器人竞争，在虚拟的火星任务中获得一席之地。该项目分析了竞争和协作对机器人关键控制概念的学习和保留的影响。该项目将制作一款低成本的教育电脑游戏，让更多的美国中学生和高中生可以通过机器人模拟游戏学习生物力学、数学、物理和计算机算法，而不是目前的机器人硬件方法。该产品为美国教育系统提供了一种解决方案，可以快速吸引那些可能会拒绝的STEM学生。STEM领域的专业可以带来高薪工作，这反过来又可以提高美国的生产力和经济前景。该项目的教育技术创新是一个基于云的机器人进化游戏引擎，学生作为一个群体可以进化双足行走机器人的行为。机器人行为的进化发生在三个层面：1)微观尺度，个体或小群体的学生进化出单一的机器人行为；2)中等尺度，群体交换个体行为，形成一个行为序列，以迎接游戏关卡的挑战。（例如，将一系列行为与站立、行走、跳跃和转身联系起来，以清除一段障碍并达到奖励目标），3)宏观尺度，即社区通过重用和采用最有效的行为、丢弃失败的行为和创造新的游戏关卡挑战而进化出一系列累积的行为。本项目特别研究了通过调整奖励激励、竞争动机、新奇性和惊喜性以及挑战等级难度对个体和群体学习的影响。在比赛过程中，游戏收集有关学生学习水平和机器人表现的数据，以回答两个具体问题：a)学生通过小团体或大团体合作/比赛学得更多，b)学生更参与小型或大型众包环境。

This SBIR Phase II project studies the problem of how to engage students in STEM (Science Technology Engineering and Mathematics) education through game-based learning in a Virtual Robotics computer game. In the game, students design, assemble and train virtual walking robots to compete against their peer robots to secure a seat on a virtual mission to Mars. The project analyses the effects of competition and collaboration on the learning and retention of key robotic control concepts. The project will produce a low-cost educational computer game, where far more U.S. middle schools and high schools students can learn about biomechanics, math, physics and computer algorithms through a robotic simulation game than is possible with current robot hardware approaches. The product provides the American educational system a solution to quickly engage students in STEM who might otherwise turn away. STEM field professions can lead to high paying jobs, which in turn can improve productivity and the economic landscape for America. The project's education technology innovation is a cloud-based robot evolution game engine where students acting as a crowd can evolve the behavior of walking bi-pedal robots. The evolution of robot behavior happens on three levels: 1) the micro scale where an individual or small group of students evolves singular robot behaviors, 2) the mid-scale where groups exchange individual behaviors to form a behavior sequence to meet game level challenges. (For example, link a sequence of behaviors to stand, walk, jump, and turn to clear a section of obstacles and reach the prize goal), 3) the macro scale where the community evolves a cumulative set of behaviors through re-use and adoption of most efficient behaviors, discarding failed behaviors and creating new game level challenges. In particular, this project studies how individual and group learning is affected by adjusting the reward incentive, competitive motivation, novelty and surprise, and challenge level difficulty. During competition, the game gathers data regarding level of student learning and robot performance to answer two specific questions: a) do students learn more through small group or large group collaboration/competition, and b) are the students more engaged in small or large scale crowd-sourced environments.