Sandeep Chinchali
University of Texas at Austin, TX·Engineering·Early-career
Fleets of robots now work in warehouses, factories, hospitals and on roads. They produce huge streams of camera and sensor data that must be processed in real time.
- Fields
Engineering
Computer Science
- Topics
cloud robotics
networked robotics
robot fleets
- Grants
- 2 known grants, through Jun 2031
What they work on
- Develop collaborative inference algorithms that use a Markov Decision Process to decide whether and where a robot should offload computation to the cloud, given network delay, power and server load.
- Design statistical data-sampling algorithms that estimate the value of uploading new robot training data compared with its labeling and training costs.
- Learn compressed representations of video and LiDAR sensor data that preserve machine-learning inference accuracy rather than human visual quality, reducing network congestion.
- Build decision-theoretic frameworks that trade off on-device edge computation and cloud inference under latency and bandwidth limits while optimizing task accuracy and compute cost.
- Create distributed data collection and multimodal retrieval algorithms that find informative and safety-critical edge cases in petabytes of robot fleet data under bandwidth and storage limits.
- Develop multimodal generative models and digital twins that synthesize realistic network latency and throughput traces to stress-test networked robots under rare failures and adversarial conditions.
More about the research
A robot faces a constant trade-off. Small models it can run onboard are fast but less accurate. Large models in the cloud are more accurate but add delay, use up network bandwidth, and raise privacy concerns. The work develops methods that let robots decide when to ask the cloud for help, what data to send, and how to adapt as network conditions change. Both projects share this theme of balancing on-robot and cloud computing. One focuses on wireless swarms over next-generation (NextG) networks. The other builds broader foundations for lifelong learning in robot fleets.
Why it matters
Smarter sharing of work between robots and the cloud could make robot fleets in warehouses, hospitals, factories and self-driving vehicles more reliable, scalable and safe. Sending only task-relevant, compressed data could reduce congestion on next-generation wireless networks and better protect privacy for people around these robots.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Shuangqing Wei
Louisiana State University, LA·Engineering
Generate a synthetic video dataset of tracheostomy-dependent infants to train privacy mechanisms, identity separation and recombination systems, and classifiers when real patient video is scarce.
Hulya Seferoglu
University of Illinois Chicago, IL·Engineering·Early-career
Characterize the cost-benefit trade-offs of coded computation, which adds redundancy to tolerate slow or failed workers, in practical edge computing systems.
Chiwei Yan
University of California-Berkeley, CA·Engineering
Develop state-aware demand control models that use real-time vehicle and rider movement to shape demand toward ride-pooling in autonomous fleets such as robotaxis, robovans and urban air mobility.
The school
University of Texas at Austin: 49 labs on PI Pages, $1.1B a year on research, 871 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.