Research Interests and Approach:

My research focuses on the design, control, and navigation of autonomous vehicles in complex physical environments, especially aquatic environments. I aim to design novel autonomous robots to investigate societally impactful problems often associated with water bodies. For example, I design surface vessels to provide new mobility to coastal cities; I develop underwater robots to monitor marine environments such as early detection of harmful algal blooms (HABs); I build and program robot swarms such as fleets of boats for new tasks such as self-assembly on the water, and underwater exploration. Moreover, always inspired by the surprising behaviors and incredible capabilities of animals, I am interested in developing biologically inspired robots and using robots to explain biology. My research approach covers a wide range of disciplines including but not limited to robotics, oceanography, computer science, biology, urban science, and artificial intelligence.

Some of the selected projects are listed here:

Multi-robot Coordination on the Water A fleet of autonomous surface vehicles can be assembled to create on-demand infrastructure, such as bridges, landing platforms, and fences, to facilitate surface activities. Due to the highly nonlinear and strongly coupled dynamics and the consistent disturbances, it is always a challenge to create autonomous self-assembly systems on the water. We are developing a group of programmable miniature robotic boats (miniboats) that can self-assemble into desired shapes and reconfigure to different shapes on the water in a distributed manner. 

Autonomous Boats for Urban Transportation Many urban cities currently have traffic problems as well as environmental problems.  Self-driving cars could mitigate these urban problems in the future. For coastal cities, we believe driverless boats (Roboat) could be an alternative to the future of transportation. Roboat can provide numerous functions for the city, such as autonomous garbage collection, on-demand delivery goods service, water quality monitoring, and transporting passengers through the city. We designed the world's first fleet of autonomous boats for urban cities and developed their control, perception, and navigation algorithms in challenging urban environments. More details.

Robust Model-based Control in Complex Physical Environments Although feedback control has been widely used in quite a few systems, it is still difficult to control robots in complex environments, such as rainy and snowy roads, waters, mining areas, rough mountains, etc. I develop robust model-based controllers such as model predictive controllers, sliding mode controllers, and active disturbance rejection controllers for robots in complex environments,  and validate these robust controllers in physical environments with robots.

Deep Reinforcement Learning Control Many model-based controllers are computationally expensive, and the performance of these controllers relies heavily on knowledge of system dynamics and uncertainties. Deep Reinforcement Learning (DRL) could be viable to control robots in dynamic environments. We design DRL controllers for autonomous vessels and compare their performance with an advanced nonlinear model predictive controller (NMPC) in real environments. Our primary results indicate that DRL controllers have lower tracking errors and offer better disturbance rejection than NMPC.

Automatic Computational Design of Robots Previously, it often took months or even years to produce suboptimal designs for autonomous vehicles such as autonomous underwater vehicles (AUVs) and unmanned aerial vehicles (UAVs), largely dependent on human experience. We are developing an automatic computational pipeline for jointly designing a robot’s shape and controller. Using our pipeline, engineers will be able to automatically search the constructed design space to create optimal designs for their requirements within a matter of hours.

Fish-like Underwater Robots  It is very challenging for underwater robots to survive and perform tasks autonomously. However, fish have survived in water for hundreds of millions of years and have evolved efficient propulsion and sensing systems. We designed fish-inspired underwater robots to study the mechanisms of efficient swimming, robust perception, and fish schooling. Based on these discoveries, we further develop bioinspired control, perception, and coordination systems for the next generation of underwater robots.

Electrocommunication Underwater communication is still challenging for underwater robots that typically have stringent power and size constraints. Weakly electric fish in nature can communicate electrically in the water by generating electrical pulses through a group of special muscle cells at the base of their tails. Inspired by electrocommunication, we invented a robust and portable artificial electrocommunication system and further built its networking system for underwater robots. We also investigate that electrocommunication systems can also be used for obstacle detection.  These bioinspired sensing and communication systems could be good supplements to sonar and acoustic communications.    

Artificial Lateral Line Sensing Perception through the water is difficult because of the low visibility of seawater. While in nature, fishes have a sensory organ called lateral line that can detect movement, vibration, and pressure gradients in the surrounding water. Lateral line plays an essential role in orientation, predatory behavior, defense, and social schooling. Inspired by the fish's "third eye", we designed artificial lateral line systems (arrays of pressure sensors around the robot's body) to investigate the sensory mechanism of lateral line. We further developed theories and algorithms to achieve underwater localization and neighboring state estimation using the developed artificial lateral line system. These artificial lateral line systems can enrich the current sensing package of underwater robots.