On the Evolution of Behaviors through Embodied Imitation

Abstract

This article describes research in which embodied imitation and behavioral adaptation are investigated in collective robotics. We model social learning in artificial agents with real robots. The robots are able to observe and learn each others' movement patterns using their on-board sensors only, so that imitation is embodied. We show that the variations that arise from embodiment allow certain behaviors that are better adapted to the process of imitation to emerge and evolve during multiple cycles of imitation. As these behaviors are more robust to uncertainties in the real robots' sensors and actuators, they can be learned by other members of the collective with higher fidelity. Three different types of learned-behavior memory have been experimentally tested to investigate the effect of memory capacity on the evolution of movement patterns, and results show that as the movement patterns evolve through multiple cycles of imitation, selection, and variation, the robots are able to, in a sense, agree on the structure of the behaviors that are imitated.

Experiments in artificial culture: from noisy imitation to storytelling robots

“…Second, we observe that behaviours adapt to be easier to learn, i.e. better ‘fitted’ to the sensorium and morphology of the robots [ 21 ], a result which appears to mirror the findings of Kirby et al [ 22 ], that artifically evolved languages evolve to be easier to learn.

Figure 2 Trajectory plot: two-robot movement-meme evolution in which all copied memes are stored, and meme selection is random with equal probability.

…”

“…In figure 3 , we see one cluster of 12 closely related memes. This result suggests the intriguing conclusion that forgetting may be a significant collective trait in behavioural evolution [ 21 ], and might also be related to what is referred to as conformist social learning, in which learners are more prone to act as others do [ 23 ].

Figure 3 Behavioural evolution map following a four-robot experiment with limited memory; each robot stores only the most recent five learned behaviours.

…”

“…Note one cluster of 12 closely related memes (55), (60–63), (67–70) and (72–74) all descended from (1). Adapted from Erbas et al [ 21 ]. …”

The development of a robust symbolic communication system for robots via embodied iterated imitation

“…In addition, we are able to examine the effects of iterated imitation cycles by comparing the results of imitation games with and without iterated imitation cycles on robots that have the same memory size. The effects of different memory architectures on the evolution of behaviors are further discussed in Erbas, Bull, and Winfield (2015).…”

Machine Ethics and the Future of Artificial Intelligence

“…In almost half of the cases, the A robot hesitated who it should save as the first one until both H-machines fell into the hole. From his experiment, Winfield concluded that robot ethics can be quite successfully substituted by machine learning [4]. The more we supply the memory of the machine with the data from the practice, the more the robot will be able to evaluate the impacts of its decision, without it being clear in advance what it will do.…”