Robotics is a natural complement to computational neuroscience. Neural models that are embodied in real world devices receive more varied inputs than simulated models, and the consequences of an embodied model's actions are less predictable --- what some dismiss as the inherent "noisiness" of the real world may in fact be a key component in how biological neural systems learn. This is one of the reasons why I feel that neural models that have physical bodies are more convincing than those that are implemented in simulation. And neuroscience models can in turn help build more capable robots. A better understanding of how animals learn motor control and cognitive abilities will produce a new set of tools for building robots that need those same skills.

My work is part of the Brain-Based Device (BBD) program here at The Neurosciences Institute. BBDs are large-scale neural models with an emphasis on realistic neuroanatomy and neurophysiology. A BBD's simulated nervous system is embodied in a robotic device, which interacts with the environment while engaged in a behavioral task.

I have been building, testing, and analyzing the Darwin X and XI BBDs, which have models of MTL plus selected cortical areas (~100K simulated neuronal units, 1.2M synapses) and perform maze-navigation tasks. Activity in the simulated hippocampus has been shown to be place specific, journey dependent and multimodal in its responses. Analysis of the neural pathways leading to place activity have pointed to differential involvement of the perforant path versus the trisynaptic loop in early versus late training place fields, and in journey-dependent versus -independent place fields. When lesioning sensory streams to the simulated MTL, changes in the pathways leading to place activity during the lesion are larger in the entorhinal cortex than in the hippocampus, which is consistent with pattern-completion theories of hippocampal function.

The work with Darwin X and XI has shown that BBDs simulating anatomical and physiological details of the MTL and surrounding regions can support the formation of spatial memory, episodic memory, and associative memory. The results using these models may have heuristic value in analyzing findings from studies of behaving animals.

Previously I have worked in the area of AI robotics, investigating issues of reliable landmark detection using predictive sensor models, symbol grounding on robotic agents, and route communication between robots.

•  JG Fleischer and JL Krichmar.   Sensory integration and remapping in a medial temporal lobe model during maze navigation by a brain-based device.  Journal of Integrative Neuroscience,  6(3):403-431,  2007.

•  DA Nitz, WJ Kargo, and JG Fleischer  Dopamine signalling and the distal reward problem.   Neuroreport,  18(17):1833-1836,  2007.

•  JG Fleischer, JA Gally, GM Edelman, and JL Krichmar.  Retrospective and prospective responses arising in a modeled hippocampus during maze navigation by a brain-based device.  Proc Nat Acad Sci USA,  104(9):3556-3561,  2007.

•  JL Krichmar, AK Seth, DA Nitz, JG Fleischer, and GM Edelman.  Spatial navigation and causal analysis in a brain-based device modeling cortical-hippocampal interactions.  Neuroinformatics,  3(3):197-222,  2005.

•  JG Fleischer.  Neural correlates of anticipation in cerebellum, basal ganglia, and hippocampus.  In, MV Butz, O Siguard, G Baldassarre, G Pezzulo (Eds.), Anticipatory Behavior in Adaptive Learning Systems: From Brains to Individual and Social Behavior,  Lecture Notes in Artificial Intelligence,  vol. 4520,   pp. 19--34,  2007.

•  JG Fleischer, SR Marsland, and JL Shapiro.  Sensory Anticipation for Autonomous Selection of Robot Landmarks.  In, MV Butz, O Siguard, P Gerard (Eds.), Anticipatory Behavior in Adaptive Learning Systems: Foundations, Theories, and Systems,  Lecture Notes in Artificial Intelligence,  vol. 2684,  pp.201--221,  2003.

•  JG Fleischer, B Szatmáry, DB Hutson, DA Moore, JA Snook, GM Edelman, and JL Krichmar.  A neurally controlled robot competes and cooperates with humans in Segway Soccer.  In Proceedings of the IEEE International Conference on Robotics and Automation,  pp.3673--3678,   2006.

•  B Szatmáry, JG Fleischer, DB Hutson, DA Moore, JA Snook, GM Edelman, and JL Krichmar.  A Segway-based human-robot soccer team.  In Proceedings of the IEEE International Conference on Robotics and Automation, pp.4436--4438,  2006.

•  JG Fleischer.  Imitation is not enough for lexicon learning  In From Animals to Animats VIII. Proceedings of the Eigth International Conference on Simulation of Adaptive Behavior,  pp.477--486,  2004.

•  JG Fleischer and SR Marsland.  Learning to autonomously select landmarks for navigation and communication. In From Animals to Animats VII. Proceedings of the Seventh International Conference on Simulation of Adaptive Behavior,  pp.151--160,  2002.

•  JG Fleischer and UDF Nehmzow.  Towards robots that give each other navigational directions: Learning symbols for perceptual categories.  In Towards Intelligent Mobile Robots 01. Proceedings of the 3rd British Conference on Autonomous Mobile Robots,  Technical Report UMCS-01-4-1,  University of Manchester Computer Science Department,  2001.

•  JG Fleischer and WO Troxell.  Biomimicry as a tool in the design of robotic systems. In Proceedings of the 3rd International Conference on Engineering Design and Automation, 1999.

•  JG Fleischer, JA Gally, GM Edelman, and JL Krichmar.  Different neural pathways lead to journey-dependent and journey-independent place cell activity in an embodied model of hippocampus.  Society for Neuroscience Annual Meeting,  2007.

•  Route Communication Between Mobile Robots Using Adaptive Landmark Symbols.  Ph.D. thesis, University of Manchester (U.K.), 2004.

•  Biomimetic design of a cooperative mobile robot system for a foraging task, M.S. thesis, Colorado State University, 1999.