136879 research outputs found
Sort by
Comparison of Evolved Memory Mechanisms in Neural Networks
No full textCognition is typically defined by the ability to develop knowledge and understanding via acquisition and processing of sensory information. Memory is a key functional requirement for cognition to establish a basis from which elevated processing is possible. There are three basic ways to implement memory using neural networks: (1) RNN : Recurrent Neural Networks, (2) TDNN : Time-Delayed Neural Networks (feed-forward), and (3): FFWD: an external marker dropper/detector network (feed-forward). In this thesis, we ask which of the three mechanisms is the easiest to evolve? To answer this question, we set up a simple ball catching task where two balls fall from above at different speeds, and an agent at the bottom has to catch the balls using range sensors. Depending on the relative speed of the balls, sometimes the slow ball will go out of sensor range; thus to catch the fast ball first then remember to catch the second (slow) ball, memory is required. We used the Neuroevolution of Augmenting Topologies (NEAT) algorithm to evolve all three types of memory mechanisms, where not only the connection weights but also the network topologies are evolved. Our results show that the FFWD mechanism is the fastest to evolve a successful controller, followed by TDNN and RNN . Among the feedforward topologies, we also found that FFWD is more robust than TDNN to certain static environmental perturbations (the relative speed of the balls), but less robust to sensory noise. We further analyze the results to probe the possible reasons for differences in evolutionary speed, as well as how different emerged behaviors influence the relationships. These results show that a simple reactive agent could quickly evolve a rudimentary form of memory through depositing and detecting external markers, long before other internalized memory mechanisms can evolve. However, the effectiveness of such elementary memory is limited. Overall, these findings shed light on the evolutionary route toward memory in cognitive agents
Fueling the mind, feeding the world: Communication in agriculture - Communicating pleasantly and professionally (COM06)
Get PDFFor additional information, visit the Texas A&M University Science Communications Lab at https://scicomm.tamu.edu/.This packet contains instructional materials and online modules prepared for Fueling the mind, feeding the world: Communication in agriculture - Communicating pleasantly and professionally (COM06). It includes curriculum, PowerPoint slides, activities, handouts, grading considerations, and notes for instructors. These materials were created as part of the USDA Grant entitled "Fueling the Mind, Feeding the World: Enhancing Communication and Decision-Making Skills of Secondary Agricultural Education Students."
MODULE OVERVIEW: This module introduces how to communicate pleasantly and professionally with examples based in agriculture. Although much of the information shared may seem obvious, brainstorming how to apply these tactics in various business communication modes is often highly beneficial to students.Secondary Education, Two-Year Postsecondary Education, and Agriculture in the K-12 Classroom (SPECA) Challenge Grants Program no. 2019-38414-30265 and Hatch Project No. TEX09825 from the USDA National Institute of Food and Agriculture
Amino-Acid-Derived Porous Materials for Chiral Resolution
No full textChirality is a fundamental property that refers to an object that cannot be superimposed on its mirror image via translation and rotation. In chemistry and biology, chiral molecules can conduct unique functions in a stereoselective environment. The vast and burgeoning market for enantiopure chemicals has stimulated the development of chiral materials for chiral resolution. Owing to their remarkably analogous physical and chemical characteristics, enantiomers pose a formidable challenge when it comes to high-efficiency and high-selectivity detection and differentiation. Therefore, novel homochiral porous materials are proposed to carry out highly efficient and selective resolution of enantiomers. In this research, natural amino acid is used as the primary chiral source because of their high accessibility, cost-effectiveness, and low toxicity. I developed a metal organic nanotube that is constructed from an amino-acid-derived ligand. The crystallographic structure is characterized by single crystal X-Ray diffraction and refined to give a unique structure possessing one-dimensional nanochannels. Additionally, the material is utilized to conduct chiral resolution on substrates to evaluate its performance. As a result, high enantioselectivity and high efficiency were found in solution-based adsorption tests. In conclusion, this research has promoted the variety of ligand geometry in construction of novel metal organic nanotube structure. The pore environments in the as-synthesized nanotube have profoundly influenced its performance in chiral resolution. Above all, this research probed the intricate relationship between structural diversity and downstream applications, proffering innovative insights to the knowledge of porous chiral materials for chiral resolution
Integrating Automation in Refueling: A Proof-of-Concept for a Universal Car Refueling System
No full textThis paper details the development of an automated refueling system, universally applicable and designed to seamlessly integrate with existing fuel stations without significant modifications. Utilizing the Kinova MICO 6 DOF robotic arm with an additional custom designed pneumatic end effector and the Intel RealSense D456 vision system, the project presents a proof-of-concept that aims to minimize the need for manual labor in extreme weather conditions. The combination of end-effectors is the key that allows for the universal applicability of this system and data from distinct sensors are effectively integrated to facilitate robust autonomous fuel dispensing. This proof-of-concept focuses on laying a versatile foundation for future enhancements rather than the development of a commercial system that addresses immediate cost reduction or safety compliance. Simulations for system validation are planned to be employed on ROS and Gazebo for effective validation of the concept, although currently the control system roadmap and effectiveness of each subsystem is verified individually, and the system can be said to be theoretically plausible. The culmination of this work establishes a groundwork for subsequent optimization towards cost-efficiency and broader applicability, promising a transformative leap in automated car refueling. Imminent steps include testing the control system integration, advancing the system with a tailor-made 6 DOF robotic arm that facilitates no constraints on the cars or fuel stations, and integrating the power and customer communication subsystems to realize a fully operational model capable of universal deployment across petrol stations leading to real world testing
