PhD Research Site of James Solberg
|
PhD Research Site of James Solberg |
|
From September 2003 through December 2007 I was a PhD studend in the Laboratory for Intelligent Mechanical Systems (LIMS) and the MacIver Laboratory at Northwestern University. This site is an overview of my PhD research, which included three main areas:
These same three areas comprise the three parts of my PhD Thesis: "Motion and Sensing in Electrosensory Systems"
Weakly electric fish use an electric field to sense their environment, enabling them to hunt at night and in turbid water. Decades of experimental and theoretical work with this group of organisms have made them a tractable model system for understanding sensory processing in vertebrates. Using idealized models of the fish's body, electric field, sensory receptors, and fluid mechanics, we investigate the structure of the active field emission and density variations in sensory receptors across the body surface, and how these properties relate to prey capture trajectories. We find that the structure of the fish's self-generated electric field provides a more regular volume for detection of prey than when the field is flipped rostro-caudally. We show that regional pecializations of sensor density on the fish's body result in better prey localization performance than a variety of other possible sensor layouts. Using an optimal control framework, we find that a subset of measured prey capture trials are both mechanically optimal and reduce sensory uncertainty faster than trials which are not mechanically optimal. The use of abstract models enables us to examine high-level questions about sensory and mechanical complementarity in animals, and can provide insight into how these two dimensions of animal life have coevolved to achieve the sophisticated behaviors they enable.
We explore the capabilities of a robotic sensing system designed to locate objects underwater through active movement of an electric field emitter and sensor apparatus. The system is inspired by the biological phenomenon of active electrolocation, a sensing strategy found in two groups of freshwater fishes known to emit weak electric fields for target localization and communication. An analytical model for the observation of simple targets is used to qualitatively predict some characteristics of the sensor including the detection distance as a function of sensor noise. We characterize the performance of the robot using different automatic electrolocation controllers, objects, and water salinity. We demonstrate successful electrolocation both in the conditions in which it is naturally bserved, in low conductivity water, as well as in conditions in which it is not observed, in water of ocean salinity. For the electrolocation experiments using an active controller based on a probabilistic sensor model, the overall median number of steps needed to locate the target is 6.7, and the median estimation error is 1.9 mm.
Full project summary: Active Electrolocation for Underwater Target Localization [ Adobe PDF ]
Here is the corresponding paper from ICRA 2007: Solberg, J.R., K.M. Lynch, and M.A. MacIver. Robotic Electrolocation: Active Underwater Object Localization with Electric Fields. in IEEE International Conference on Robotics and Automation. 2007. Rome, Italy. [Adobe PDF]
See the on-line article from New Scientist: Electric fields could give subs 'fish-like' sense
"Uncertainty Driven Motion Plan" Addresses the basic question of "How can I move my sensors to minimize uncertainty?". Here, I mainly look at linear systems with Gaussian and white noise (thus, can use Kalman filter). Downloadable versions: Adobe PDF | MS PowerPoint
"Motion Planning for Active Sensing" Begins with brief introduction to Bayesian filters and information metrics. Then, I introduce a brute-force method of sensor-based motion planning for a fish prey strike. Downloadable versions: Adobe PDF | MS PowerPoint
"Electrolocation" I explain the basics of the electrolocation task that applies for both robots and weakly electric fish. Downloadable versions: Adobe PDF | MS PowerPoint
"Robotic Electrolocation: Active Underwater Target Localization with Electric Fields" I describe the hardware, software, and performance of a robot capable of locating underwater targets using electrosense and an active control scheme. Downloadable version: MS PowerPoint
In December of 2005 I presented my PhD proposal to my committee. Downloadable versions:
Solberg, J.R. and P.M. Ferreira, Inverse Kinematics of Robotics and Its Applications of Symbolic Computing. 1996, Wolfram Research Inc. Technical Reprort: Champaign, IL.
Solberg, J.R. and N.R. Miller, Regulation of the Liquid-Mass-Fraction of the Refrigerant Exiting an Evaporator., in Technical Report TR-159 for the Air Conditioning and Refrigeration Center. 2000: Urbana, IL. p. 174.
Solberg, J.R., N.R. Miller, and P.S. Hrnjak. A Sensor for Estimating the Liquid Mass Fraction of the Refrigerant Exiting an Evaporator. in SAE 2000 World Congress. 2000. Detroit, MI.
Solberg, J.R. and R.M. Smith, Closed-Loop Control of Functional Electrical Stimulation for Human Biped Locomotion in Paraplegic Patients, in Faculty of Health Sciences. 2001, Sydney University: Sydney, Australia.
Solberg, J.R., Sensor-Based Control in Weakly Electric Fish with Applications to Robotics, in Technical Report for The Laboratory for Intelligent Mechanical Systems. 2005: Evanston, IL.
Solberg, J.R., K.M. Lynch, and M.A. MacIver. Robotic Electrolocation: Active Underwater Object Localization with Electric Fields. in IEEE International Conference on Robotics and Automation. 2007. Rome, Italy.
Solberg, J.R., K. M. Lynch, and M. A. MacIver. Active electrolocation for
underwater target localization. Submitted to the International Journal
of Robotics Research, 2007.
Solberg, J.R., K. M. Lynch, and M. A. MacIver. Optimal Sensing in Weakly
Electric Fish To be submitted to PLoS Computational Biology.