Sandeep Chichali, 16 Oct 2019
Sandeep Chinchali, a Caltech alum and PhD students at Stanford, will visit Caltech onn 16 Oct (Wed). If you would like to meet with him, sign up here.
- 11 am: seminar
- 12 pm: lunch with Richard (and Joel, if he is around)
- 1 pm: Sumanth (Ann Lounge, level 3)
- 1:30 pm: Open
- 2:00 pm: Open
- 2:30 pm: Francesca
- 3 pm: CDS tea (optional)
- 3:45 pm: RMM group meeting (you are welcome to come if you like)
Distributed Perception Between Robots and the Cloud: A Learning-Based Approach
Sandeep Chinchali, PhD candidate, Stanford
16 Oct (Wed), 11a-12p, 121 Annenberg
Today’s robotic fleets are increasingly facing two coupled challenges. First, they are measuring growing volumes of high-bitrate video and LIDAR sensory streams, which, second, often leads them to use increasingly compute-intensive models, such as deep neural networks (DNNs), for downstream perception or control. To cope with these intertwined challenges, compute and storage-limited robots, such as low-power drones, can offload data to central servers (or “the cloud”), for more accurate real-time perception as well as offline model learning. However, cloud processing of large robotic sensory streams introduces acute systems bottlenecks ranging from network delay for real-time inference, to cloud storage, human annotation, and cloud-computing costs for offline model learning.
In this talk, I will present learning-based approaches for robots to improve model performance with cloud computing, but with minimal systems cost. For real-time inference, I will present a deep reinforcement learning based offloader that decides when a robot should exploit low-latency, on-board computation, or, when highly uncertain, query a more accurate cloud model. Then, for continual learning, I will present an intelligent, on-robot sampler that mines real-time sensory streams for valuable training examples to send to the cloud for model re-training and specialization. Using insights from months of field data and experiments on state-of-the-art embedded deep learning hardware, I will show how simple learning algorithms can allow robots to significantly transcend their on-board sensing performance, but with limited cloud communication cost. Finally, I will conclude with future directions in data-driven networked control, informed by my industry collaborations.
Bio: Sandeep Chinchali is a final-year Computer Science PhD candidate at Stanford, advised by Marco Pavone and Sachin Katti. Previously, he was an early engineer at Uhana, a Stanford startup working on data-driven optimization of cellular networks, now acquired by VMWare. His research on cloud robotics and data-driven control of wireless systems has led to successful proof-of-concept trials with major cellular network operators, and was a finalist for best student paper at Robotics: Science and Systems 2019. Prior to Stanford, he graduated from Caltech, where he worked on applications of formal methods in robotics.