This book constitutes the refereed proceedings of the Third Workshop on Human Behavior Understanding, HBU 2012, held in Vilamoura, Portugal, in October 2012. The 14 revised papers presented were carefully reviewed and selected from 31 submissions. The papers are organized in topical sections on sensing human behavior; social and affective signals; human-robot interaction; imitation and learning from demonstration.

A team of relatively simple robots may achieve a complex goal more effectively than a single complex robot if a proper design paradigm is used. Multirobot systems have a wide application area from search and rescue operations in disaster areas and planetary exploration to soccer playing. Robot soccer is a good platform to test and develop multi-robot applications because it has some physical limitations such as limited and noisy sensorial information and noisy actuators as in the real life and it also has a highly dynamic environment. The ultimate goal of winning the game should be decomposed into a sequence of sub-goals and proper sequences of actions for achieving the subgoals should be selected and refined through execution. In order to be able to select proper actions at a time, the system should be able to evaluate the current state therefore some metrics that gives quantitative information about the environment are needed. This book presents a set of metrics calculated from positions of robots and ball on the field, and a statistical method for evaluating their usefulness. A task allocation algorithm built on top of those metrics is also presented.