Multi-agent trajectory simulation