National Science Foundation
Exploration and Exploitation in Actuated Communication Networks
University of Southern California (USC): Urbashi Mitra, Shrikanth Narayanan and Gaurav Sukhatame
Massachusetts Institute of Technology (MIT): Franz Hover
Northeastern University (NEU): Milica Stojanovic
Mei Yi Cheung, Eric Gilbertson and Brooks Reed
The oceans cover 71% of the earth's surface and are integral to climate regulation, nutrient production, oil retrieval, and transportation; yet they represent one of the least explored frontiers. It can be reasonably argued that we know more about the solar system and space, than we do about the parts of planet Earth covered by water. Future scientific and technological efforts to achieve better understanding of oceans and water-related applications will rely heavily on our ability to jointly consider communications, actuation and sensing in a unified system that includes instruments, vehicles, human operators and sensors of all types.
The goals of this project are to design networking tools for mobile underwater networks, develop novel navigation mechanisms for communication-constrained autonomous underwater vehicles and ultimately to integrate sensing and classification to provide solutions for the exploration-exploitation tradeoff. We seek to address (a) mobile underwater network design; (b) AUV routing with communication constraints; and (c) integrating sensing to design exploration/exploitation algorithms.
We will explore fundamental properties of underwater actuated acoustic communication networks in order to design and analyze high-performance exploration and exploitation algorithms. We shall provide implementable algorithms which will endeavor to achieve the fundamental limits, as well as performance analyses for our methods. The research outcomes will impact: robust distributed resource-constrained classification, multi-modal classification methods, sparsity in pattern recognition vehicle routing, stochastic motion planning, spectral graph theory, SDP relaxation, routing for actuated networks, action-dependent communication, active classification, localization and node discovery methods, medium access for actuated networks, power control algorithms.etc.
The broader impacts of the proposed research activity include: (i) training of new information technology professionals and scientists with expertise in interdisciplinary research; (ii) training of technology professionals and scientists in both theory and experiment; (iii) the development of methods with application to science, security, and industry in the areas of environmental monitoring, aquatic eco-system analysis, ocean accident remediation, surveillance for defense applications, homeland security, oil and gas, aquaculture, geological and oceanographic science, marine biology, etc.; and finally (v) active engagement of the PIs to work with under-represented women and minorities in their research program, leveraging existing efforts and prior success.
The Hover research group aims to enable truly dynamic multi-vehicle missions underwater through integrated design of control and communication. In particular, we develop advanced control analysis and synthesis methods for feedback control that relies on the lossy and rate-limited acoustic channel. Field tests with autonomous underwater and surface vehicles and acoustic modems serve to validate these designs as well as characterize their real-world performance.
Multi-vehicle dynamic pursuit
This work experimentally demonstrates a prototypical multi-vehicle tracking and pursuit system, where two vehicles jointly estimate the location of a mobile target, using entirely acoustic means. The control cycle includes range measurements and two-way communications between the vehicles. Estimation and control are tightly coupled in this mission, allowing us to study the effects of packet loss, quantization, delays and scheduling on the frequency response of the integrated closed-loop system. Experimental results show that tracking bandwidths of roughly half the Nyquist rate are achievable.
Vehicle-based sampling of the ocean traditionally has involved lawnmower or yo-yo surveys that allow for reconstruction of quasi-static fields of interest. With recent improvements in vehicle and ocean modeling technology, it is desired to track dynamic features such as fronts and plumes at higher temporal and spatial resolutions than previously possible. We are developing methods for multi-vehicle tracking and pursuit of ocean features that decompose spatial and temporal variations, as well as allow for application of advanced control design. The concept is that locally linear behavior of an ocean process admits powerful network-based control techniques on short time scales. Our LTI framework couples global structure of the ocean process (obtained via ocean models and system identification) to vehicle positioning. To start, we have combined this framework with a control technique that is robust to packet loss. Simulations using ocean models show that the resulting integrated, collaborative system outperforms methods that act on local information alone.
Channel Describing Functions
Many multi-agent robotic applications such as pursuit and monitoring depend heavily on wireless communications, whose performance can impact the whole system. We derive a closed-form describing function for a combined quantizer, binary erasure channel, and linear filter decoder for a SISO system; this combination is a fundamental model in wireless feedback control, and has wide practical relevance. Describing functions identify limit-cycles in nonlinear systems, and thus the result allows us to rapidly characterize the dynamic precision of wirelessly-controlled systems. The key steps in our derivation can be applied to other fundamental systems, and we derive describing functions for systems where the linear filter element is replaced with zero-output and hold-output decoders. Analysis of these systems using the describing function technique is orders of magnitude faster than Monte Carlo simulation.
Multi-Armed Bandit Adaptive Relay Positioning
The performance of the acoustic channel depends on local water properties, bathymetry and other environmental variables. In shallow-water and man-made environments, multipath interference results in significant variability in space. A key insight lies in recognizing that acoustic nodes can learn about and exploit the statistics of the channel at their current location while receiving and transmitting. For the purpose of improving acoustic communications during a multivehicle mission, we consider the case of deploying a mobile acoustic relay to adapt to acoustic link variations in physical space. Thus, the relay must balance searching unknown sites to gain more information, which may pay off in the future, or exploiting already-visited sites for immediate reward. This is a classic exploration vs. exploitation problem that is well-described by a multi-armed bandit framework. We formulate the multi-armed bandit for the problem of adaptive acoustic relay positioning and apply an optimal solution in the form of Gittins indices. However, for an autonomous ocean vehicle traveling between distant waypoints, the time costs of traveling between locations (switching costs) are significant. The multi-armed bandit with switching costs has no optimal index policy, so we develop an adaptation of the Gittins index rule with limited policy enumeration and asymptotic performance bounds.
Feedback Control with Packet Loss
We consider strategies for feedback control with acomms links in either the sensor-controller channel, or the controller-actuator channel. On the controller-actuator side we implement sparse packetized predictive control (S-PPC), which simultaneously addresses packet-loss and the data rate limit. For the sensor-controller channel we study a modified information filter (MIF) in a Linear Quadratic Gaussian (LQG) control scheme. Field experiments were carried out with both approaches, regulating crosstrack error in a robotic kayak using acomms. Outcomes with both the S-PPC and MIF LQG confirm that good performance is achievable.
Robust Minimum Energy Wireless Routing
As vehicle fleet sizes grow, network routing rises in importance. However, the acoustic channel is stochastic, and power is limited. This work formulates the robust counterpart of a mixed-integer optimization approach to minimum energy wireless multicast. We derive scaled power levels that account for uncertainty and allow for efficient solution via standard MILP solvers.
Kinematically-constrained Compressed Sensing
Compressive sampling and compressed sensing have offered an alternative approach to Nyquist-constrained sampling and reconstruction, by leveraging sparsity of the field variable(s) in some dictionary. When sampling is done by a physical vehicle in space, samples are expensive (in terms of time and energy) and optional. We have investigated the design of sampling projection matrices that simultaneously achieve efficient reconstruction (via the underlying coherence properties in sparse reconstruction), and are efficient for the vehicle.
Reed, B., Stojanovic, M., Mitra, U. and F. Hover, "Robust Minimum Energy Wireless Routing for Underwater Acoustic Communication Networks," IEEE Globecom 2012 Workshop on Wireless Networking and Control for Unmanned Autonomous Vehicles: Architectures, Protocols and Applications (Wi-UAV Workshop), Anaheim, CA, December 2012. (Full Text, PDF)
Cheung, M., Leighton, J. and F. Hover, "Multi-armed Bandit Formulation for Autonomous Mobile Acoustic Relay Adaptive Positioning", IEEE International Conference on Robotics and Automation (ICRA), Karlsruhe, Germany, May 2013. (Full Text, PDF)
Gilbertson, E., Reed, B.L., Leighton, J., Cheung, M. and F. Hover, "Experiments in Dynamic Control of Autonomous Marine Vehicles Using Acoustic Modems", IEEE International Conference on Robotics and Automation (ICRA), Karlsruhe, Germany, May 2013. (Full Text, PDF)
Reed, B., and F. Hover, "Tracking Ocean Fronts with Multiple Vehicles and Mixed Communication Losses", IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Tokyo, Japan, November 2013, to appear. (Abstract, PDF)
Cheung, M. and F. Hover, "Autonomous Mobile Acoustic Relay Positioning as a Multi-Armed Bandit with Switching Costs", IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Tokyo, Japan, November 2013, to appear. (Abstract, PDF)
Gilbertson, E. and F.S. Hover, "Limit Cycling in Control of Underwater Vehicles Caused by Quantized Hold-Input Binary Erasure Communication Channels," International Symposium on Unmanned Untethered Submersible Technology (UUST), Portsmouth, NH, August 2013.(Full Text, PDF)
Cheung, M. and F.S. Hover, "A Multi-Armed Bandit Formulation with Switching Costs for Autonomous Adaptive Acoustic Relay Positioning," International Symposium on Untethered Unmanned Submersible Technology (UUST), Portsmouth, NH, August 2013. (Full Text, PDF)
Reed, B.L., and F.S. Hover, "Experiments in Joint Estimation and Pursuit with Marine Vehicles through Acoustic Communication," International Symposium on Unmanned Untethered Submersible Technology (UUST), Portsmouth, NH, August 2013. (Full Text, PDF)
Cheung, M., Leighton, J., Mitra, U., Singh, H. and F. Hover, "Field Experiments in Acoustic Relay Positioning as a Multi-Armed Bandit with Switching Costs", Submitted to International Symposium on Robotics Research (ISRR), Singapore, December 2013. (Abstract, PDF)
Reed, B., Leighton, J., Stojanovic, M. and F. Hover, "Multi-Vehicle Dynamic Pursuit using Underwater Acoustics", Submitted to International Symposium on Robotics Research (ISRR), Singapore, December 2013. (Abstract, PDF)