Description: In robotic applications where privacy is important, robots could be observed or comprised. To accomplish its task, robots need to collect information that is valuable (for tracking or planning tasks), but this information can be sensitive if leaked. A privacy-preserving tracking example is shown above: a robot is tracking a panda, but the robot is compromised by an eavesdropper, who is also interested in knowing the location of the panda. As a consequence, every information known by the robot is known by the panda. To accomplish the tracking task while protecting the panda from the eavesdropper, the robot should track the panda in a precision that is no coarser than a tracking bound, but no finer than a privacy bound.

In this project, we firstly examined privacy-preserving tracking, and demonstrated several impossibility results. For example, the feasibility to achieve privacy-preserving tracking is sensitive to the robot's prior knowledge about the target's initial location; The robot’s power to achieve privacy-preserving tracking is bounded, converging asymptotically with increasing sensing power. We also examined the task of privacy-preserving planning, where the robot aims to find a plan that achieves some goal as well as constraining the information disclosed from the plan. A knowledge gap is created between the robot and the observer via the observer's prior knowledge about the robot's plan, information disclosure policy which determines how information is disclosed from the robot to the observer, and the structure of the observer's estimator. We managed to build search algorithms to search for plans and information disclosure policy jointly, while constraining the observer's knowledge.

[1] Yulin Zhang, Dylan A. Shell. You can't save all the pandas: impossibility results for privacy-preserving tracking. International Workshop on the Algorithmic Foundations of Robotics, 2016. [PDF]
[2] Yulin Zhang, Dylan A. Shell. Complete characterization of a class of privacy-preserving tracking problems. International Journal of Robotics Research, 2018. [abstract] [arXiv]
[3] Yulin Zhang, Dylan A. Shell, Jason M. O'Kane. What does my knowing your plans tell me? IROS-Cogrob, 2018. [arXiv]
[4] Yulin Zhang, Dylan A. Shell, Jason M. O'Kane. Finding plans subject to stipulations on what information they divulge. International Workshop on the Algorithmic Foundations of Robotics, 2018. [arXiv]
[5] Yulin Zhang, Dylan A. Shell. Plans that remain private even in hindsight. AAAI-PPAI, 2020. [PDF]