NIMS: 3-dimensional, aquatic & autonomous-IDEA
- Author(s): Stealey, Michael J
- Borgstrom, Per Henrik
- Singh, Amarjeet
- Jordan, Brett
- Chen, Victor
- Batalin, Maxim A
- Kaiser, William J
- et al.
During the last decade as a result of increasing concern for water resource availability, the complexity of aquatic sensing applications has increased as a result of demands for: 1) broad spatial coverage and high spatial resolution monitoring, 2) capability for resolving fine scale spatiotemporal dynamics and 3) the need for rapid system deployment with automatic operations.
Current research is aimed at the implementation of a four cabled Aquatic Networked InfoMechanical Systems (NIMS-AQ) in a kinematically redundant configuration. This configuration requires active cable tension control, which is accomplished by means of a cantilevered load cell and a PID controller. System positioning is controlled by adjusting the tension levels in each of the four cables to generate the desired net force on the end-effector. Tension configurations are not unique (due to kinematic redundancy) and the optimal configuration is found by means of a novel approach that reduces the problem to a two-dimensional linear programming optimization (in real-time). Real-time, three-dimensional control coupled with underwater sonar enables NIMS-AQ to perform precise, autonomous calibration and environmental characterization (spatial and semantic mapping). Dedicated on-board sonar ensures high fidelity monitoring of the underwater environment, greatly expediting experimental design and system setup.
Efficient, accurate system control and environmental characterization help realize an Autonomous Iterative experimental Design for Environmental Applications (A-IDEA) conjunctive network. IDEA provides a methodology for in-field adaptation of experimental design to perform detailed characterization of the spatiotemporal distribution of the observed environment. This involves an in-field adaptation in the experiment design to capture phenomena dynamics exploiting observations from prior models, iteratively executed experiments and the behavior of the underlying control processes (if known).