A new performance bound for submodular maximization problems and its application to multi-agent optimal coverage problems