DARPA challenges, RoboCup, and RoboCup@Home
Over the few last decades many diverse robot competitions and challenges - sponsored by different organizations like the DARPA (Defense Advanced Research Projects Agency), the Robot World Cup (RoboCup) Federation, the AAAI (Association for the Advances on Artificial Intelligence), and many others - have been proposed, with the aim of testing the level of development of embodied systems. These competitions have never been proposed as being scientific tests for evaluating the level of “intelligence” in such systems nor their eventual cognitive plausibility. Rather, they have been proposed to evaluate the degree of human-level abilities reached by competing teams working on different systems, platforms, and integrated architectures. As such, they hold a different status with respect to some of the proposals reviewed above. However, they have provided a valuable alternative and have the great merit of creating a community of researchers working together on integrated platforms. One of the most famous events is the DARPA Challenges, a series of competitions sponsored by the DARPA in order to promote the development of autonomous vehicles able to complete, in a limited amount of time, substantial off-road courses. These original challenges (called “Grand Challenges”) were later on substituted with different competitions, like the DARPA Urban Challenge, which extended the initial challenge to autonomous operation in a mock urban environment; and, more recently (since 2012), with the DARPA Robotics Challenge, focused on human-supervised robotic technology for disaster-response operations. Another well-known competition, a robot soccer one, known as the Robot World Cup (RoboCup), started in 1997 and has been considered an important testbed for the investigation of cooperative multi-agent systems. Such leagues started with non-humanoid robots but nowadays employ humanoid soccer robots. The stated ultimate goal of RoboCup (www.robocup.org) is described as follows: “By the middle of the 21st century, a team of fully autonomous humanoid robot soccer players shall win a soccer game, complying with the official rules of FIFA, against the winner of the most recent World Cup” (Kitano and Asada, 2000) and, by the year 2030, soccer robots are expected to play competitive games against a team of eight unprofessional players (Gerndt et al., 2015). One thing that has certainly emerged already from the preliminary competitions is the adoption in most teams of hybrid approaches combining low-level reactive systems and higher-level modules for planning and reasoning. This marriage was unavoidable since, as we have seen, certain classes of formalisms and architectures (e.g., those a la Brooks) are better suited to dealing with perceptual tasks while others (e.g., the heuristic search methods used in robots, like Shakey, see footnote 20 in Chapter 1) are able to better model high-level phenomena. More recently, the RoboCup organization has proposed a novel competition called the “RoboCup@Home” league that aims to evaluate the level of abilities of assistive robots in domestic applications. So in this case, the idea is one of fostering research on human-robot interaction instead of on multi-agent robot-robot interactions. All these challenges are important testbeds for evaluating the progress and problems of embodied systems showing integrated abilities.