Robotics Research

My doctoral research focused on decentralized multi-robot coordination and formation control, spanning from 2012 to 2016. I developed provably correct algorithms for scalable multi-robot systems that eliminate the need for central coordination, enabling robots to autonomously form complex patterns and adapt to dynamic environments.

My key contributions include: "Repeating Patterns of Mobile Robots: A Provably Correct Decentralized Algorithm" (IROS 2016), which established mathematical foundations for decentralized pattern formation; "Decentralized formation of arbitrary multi-robot lattices" (ICRA 2014), enabling flexible formation control for any desired lattice structure; "Comparison of constrained geometric approximation strategies for planar information states" (ICRA 2012), providing theoretical frameworks for efficient state representation; and "A grid-based approach to formation reconfiguration for robots with non-holonomic constraints" (European Journal of Control 2012), bridging theoretical control with practical robot constraints.

I and Prof. Jason M. O'Kane, 2015, Univ of South Carolina

Decentralized formation algorithm, ICRA 2014

Publications

Y. Song and J. M. O'Kane, "Repeating Patterns of Mobile Robots: A Provably Correct Decentralized Algorithm. I", IROS 2016.

Y. Song and J. M. O'Kane, "Decentralized formation of arbitrary multi-robot lattices", ICRA 2014.

Y. Song and J. M. O'Kane, "Comparison of constrained geometric approximation strategies for planar information states", ICRA 2012.

D. Miklic, S. Bogdan, R. Fierro, Y. Song, "A grid-based approach to formation reconfiguration for a class of robots with non-holonomic constraints", European Journal of Control 18 (2), 162-181, 2012.

Experience

Senior Algorithm Engineer	2022 – Current	Lotus Tech	+
Architected and manage a cloud-native SaaS platform on AWS supporting ADAS data processing and data-closed loops for autonomous automotive systems. Oversaw compliance, data lifecycle, infrastructure reliability, and performance optimization for ADAS workloads across hybrid cloud architectures. Built Kubernetes-based data platforms that process 10M+ daily sensor data points using MSK, Lambda, S3, and Redshift, implementing automated ML workflows for ADAS model training. Drive business development initiatives with external automotive clients—gathering requirements, shaping technical proposals, and delivering POCs to win projects and expand partnerships.
Senior Algorithm Engineer	2019 –2022	Aptiv	+
Led radar processing algorithms for Motional's Robotaxi, enhancing sensor fusion libraries with Python, C++, and in-house cloud-based toolchains. Led POC development of lane change prediction and ADAS features for BMW and Stellantis clients. Delivered error-handling APIs with ROS, ensuring ASPICE-compliant solutions.
Software Engineer	2016 – 2019	Groupon	+
Built REST APIs in Java and Scala, enabling 6M+ daily customer notifications and GDPR compliance for EMEA. Migrated Hive to Spark, boosting performance 10x and cutting data pipeline runtime by 50%+. Designed and implemented scalable data processing architectures using Apache Spark, Kafka, and cloud-based data warehouses for real-time analytics and customer insights.
Robotics Engineer Intern	2015 – 2015	Auro Robotics	+
As Motion Planning Engineer at Auro Robotics during YC Summer 2015, I developed ROS-based path planning algorithms for a driverless campus shuttle from scratch. Worked directly with co-founders in Mountain View garage, integrating GPS and LiDAR for autonomous navigation. Successfully demonstrated at Stanford campus during YC Demo Day, securing $2.1M funding from Sam Altman and investors.

Yang Song, Ph.D.