NICOP - Developmental Cognitive Robot with Life-long Learning
Abstract
ProjectAbstractThe project objective is to develop technologies necessary to enable a next-generation cognitive robot with adaptive artificial intelligence system. The system is capable of continuous, one-shot learning with small data and improves performance in real-world environments while constrained by pre-determined capability limits. This system would be able to apply existing knowledge tonew circumstances without the need of previous information or training data, and would be able to update its network in variety of applications. The Developmental Cognitive Architecture with Life-long Learning (CogL3) architecture is developed based on the developmental cognitive architecture and four principles of lifelong machine learning model. The goals are to be able toperceive its environment, learn from experience, anticipate the outcome of events, act to pursue goals, and adapt to changing circumstances. To perform continual learning, this project is designed based on the open-ended life-learning learning paradigm that learns perpetually using external data and internal goals. The CogL3 architecture demonstrates the ability to learn newtasks without losing capability on previously learned tasks, and can apply previous knowledge to novel situations; and in so doing, can develop more complex capabilities. The resulting systems is more reliable and robust when used in dynamic or unpredictable situations.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 10, 2018
- Source ID
- N629091812086
Entities
People
- Chu Kiong Loo
Organizations
- Office of Naval Research
- United States Navy
- Universiti Malaya