Zhengping Ji

Office: 190A Mellon Institute,                             Carnegie Mellon University 
Lab Phone: (517) 432-9474
Home Phone: (517) 775-1459
Email: jizhengp@andrew.cmu.edu

PREVIOUS RESEARCH WORK

Developmental networks, called Where-What Networks (WWN) [8][9][10][11][12][13][14], were designed for a general sensorimotor pathway, such that recognition and attention interact with each other in a single network. The cortex inspired neuromorphic architecture modeled three types of attention: feature-based bottom-up attention, location-based top-down attention, and object-based top-down attention. The in-place learning algorithm was used to develop the internal representation (including synaptic bottom-up and top-down weights of every neuron) in the network, such that every neuron was responsible for the learning of its own signal processing characteristics within its connected network environment, through interactions with other neurons in the same layer.

Object Classification by the Sensor Fusion of Radar and Vision Systems

A low cost, real-time object recognition system was developed in a sensor fusion framework [3][6][7][15][16]. The powerful vision algorithm is used to deal with a wide variation of targets provided by radar system. The objects of interest fall into three categories:(1) vehicle (e.g., cars and trucks); (2) pedestrians; (3) other objects having radar returns (e.g., light poles, guide rails, signs and bridges, garbage cans, road barriers). The module clusters multiple radar tracks into a single object and facilitates the threat assessment algorithms by offering object category labels. It fits into an embedded system and holds promises in industrial applications. 

Autonomous Mental Developmental Learning

This research is to advance artificial intelligence using what we call the developmental approach [2][4]. The new approach is motivated by human cognitive and behavioral development from infancy to adulthood. It requires a fundamentally different way of addressing the issue of machine intelligence. Developmental programs, such as SHM, IHDR (e.g., [17]), MILN, etc., was applied for a robot to develop its mind through the learning process. 

Autonomous Outdoor Navigation

Developed an online-learning and attention-based approach to outdoor navigation [1][5][18] in DARPA Grand Challenge, 2005. Several sensors were applied in vehicle driving to percept local environments by a combination of vision, LADAR, ultrasonic, and contact sensors. The work was applied on Team AVS and Team Crossland in the challenge events.