Sandeep Chinchali

Postdoctoral Scholar, Computer Science Department, Stanford University e-mail: csandeep [@] stanford [DOT] edu

About me

I will join the University of Texas at Austin as an assistant professor in the ECE department and a core member of UT Austin's Robotics consortium in summer 2021. I am looking for graduate students to join my research group. Please contact me if you are interested in robotics, edge computing, 5G networks, and related topics!

I recently completed my Computer Science PhD at Stanford University co-advised by Prof. Sachin Katti and Prof. Marco Pavone, where I worked on cloud and networked robotics. My PhD thesis, titled Distributed Perception and Learning Between Robots and the Cloud, uses tools from deep learning, computer networks, and data-driven control.

Previously, I was the first principal data scientist at my adviser’s startup Uhana (acquired by VMware), where I built deep-learning-based models for cellular network optimization and control, some of which were successfully piloted in proof-of-concept trials by major network operators. Prior to Stanford and Uhana, I received a BS (Honors) in Electrical Engineering and a minor in Control and Dynamical Systems from Caltech. At Caltech, I was president of the Tau Beta Pi Engineering Honors Society and worked on formal verification for robotics with Professors Joel Burdick and Richard Murray.

Selected Papers

A full list of publications and citations can be found here.

• Harvestnet: Mining valuable training data from high-volume robot sensory streams.

• Cellular network traffic scheduling with deep reinforcement learning.

• Network Offloading Policies for Cloud Robotics: a Learning-based Approach.

• Multi-objective optimal control for proactive decision making with temporal logic models.

Research

My research applies deep reinforcement learning (RL) to a variety of scheduling and control problems for cellular networks, Internet-of-Things (IoT Devices), and robotics. My publications span both robotics/AI and systems venues:

Systems:

• Learning Based Approaches to Network Offloading in Cloud Robotics (Robotics: Science and Systems 2019)

  • Finalist for Best Systems Paper and Best Student Paper

• Scheduling Traffic from IoT Devices on Cellular Networks Using Deep RL (AAAI 2018)

• Distributed Neural Network Inference Between Edge Devices and the Cloud (ACM HotNets 2018)

• Decoupling Prediction and Control Across Data Boundaries

Robotics:

• Proactive Decision Making for Autonomous Driving using Deep RL and Formal Methods (International Symposium on Robotics Research 2017)

  • Invited and Accepted to a Special Edition of the International Journal of Robotics Research (IJRR)

• Formal Methods for Autonomous Car Safety Certification (ICRA 2016)

• Towards Formal Methods for Dexterous Robotic Manipulation (ICRA 2012)

A full list of publications can be found here.

Invited Talks

I have given talks both in industry and academic conferences, including Google X, the ITU Workshop on “Machine Learning for 5G and beyond”, and Microsoft Research. Talk slides can be found here.