CPS- Synergy- Collaborative Research- Towards Effective and Efficient Sensing-Motion Co-Design of Swarming Cyber-Physical System

The overall research objective of the project is to establish and demonstrate a generic motion- sensing co-design procedure that significantly reduces the complexity of mission design for swarm- ing CPS, and greatly facilitates the development of effective and efficient control and sensing strategies. The objective of the project will be achieved through the integration of three main com- ponents: the design of cooperative control and sensing strategies, the development of Magnetic- induction-based underwater communication and localization technique, and the design of smart- material actuated biorobotic fish.

Source seeking is one of the fundamental and representative missions for swarming CPS with a wide range of practical applications. We have developed a dual-module control approach that achieves fast source seeking using minimum number of mobile robots in a static field and val- idated the approach in a multi-robot testbed using various types of mobile robots with different capabilities. This year, we focus on environmental processes that are spatial-temporal varying. A cooperative filtering scheme is developed to achieve online parameter identification of the un- known field using a swarming CPS. Source seeking algorithms are extended to accommodate the spatial-temporal varying feature of the field. To validate the proposed algorithms under realistic uncertainties and disturbances, we build a controllable CO2 diffusion field and construct a CO2 sensor grid to calibrate the field. Experiments are conducted using four mobile robots with CO2 sensors in the controllable diffusion field. The four robots successfully locate the source of the dif- fusion field while maintaining a desired formation. The online estimates of the diffusion coefficient converge in the experiments.

Collision avoidance is an important requirement in vehicle swarms. In our project, we em- ploy the collision cone approach. The collision cone approach has several advantages over other approaches (such as potential fields, navigation functions, RRTs, etc.) in that it is a reactive, on-line technique suitable for dynamic environments, and possesses strong analytical founda- tions. This enables us to determine analytical guidance laws for collision avoidance, which offer advantages over other numerical approaches, in that it can lead to computational savings on resource-constrained mobile robotic platforms. Additionally, while most collision avoidance tech- niques tend to approximate the shape of the robots and obstacles by circles/polygons, the collision cone approach, when used for planar dynamic environments, does not require any such approxi- mation. Even for objects of completely arbitrary/non-convex shapes, the collision cone approach provides analytical guidance laws that come with defined collision avoidance guarantees. The fact that such guarantees are provided even for arbitrarily shaped objects provides a larger amount of "free-space" for the robots to move within their environment. This allows the robots room to maneuver even in cluttered environments. The collision cone approach has also been used for meeting other objectives, in addition to collision avoidance. These include generating trajectories for a robot to pass through an orifice.

To provide an enabling mobile platform to verify the proposed strategies, we develop a 2D maneuverable robotic fish propelled by multi-IPMC fins. One caudal fin is used for forward swim- ming. Two pectoral fins are used for turning. A wireless control system is developed for controlling the speed and direction of the swimming robotic fish, where a Xbee device is used for receiving commands from a PC station. We have fabricated the robotic fish at Wichita State University. The total weight of the robot was 290 grams. We have tested the robotic fish in a water tank. The free swimming tests show that the robotic fish can achieve 0.5 cm/sec forward speed and 1.5 rad/sec turning speed. We have developed the dynamic model for the robotic fish and have validated the model.
To allow robotic fish to exchange messages with reliably controllable performance in the harsh underwater environment, we develop novel Magnetic Induction (MI)-based underwater communi- cation module. The contribution of this project in the past year focus on developing an analyti- cal channel model for MI underwater communication to characterize the complex underwater MI channels, especially in the shallow water with omnidirectional antennas, and developing and imple- menting the environment-aware and MI-based localization technique. This year, three propagation paths are theoretically modeled: direct path, reflected path, and lateral wave. In-lab experiments and COMSOL simulations are conducted to validate the theoretical model. It is observed that Tri-directional coil antenna can reduce the orientation effect. In addition, lateral waves take strong effect in shallow water, especially when distance is larger than the depth of transmitter and re- ceiver. Furthermore, we find that Mbps data rate in near region (less than 5m) while kbps data rate in several tens meters. For the MI-based localization, the multi-path fading-free MI channel & orthogonality of tri-coil MI antennas can provide accurate, simple, and convenient localization strategy. By using 3 coils in orthogonal planes, we can determine the positions of sensors in 3D space while only one anchor node is needed. However, the existence of the highly-conductive ob- jects may influence the received MI signal strength. Hence, we developed the environment-aware MI localization technique.