SPOTting the Blind Men and the Elephant
PI: Dr. Charles Kim, Department of Electrical and Computer Engineering, Howard University, Washington, DC
Project Sponsor: Sun Microsystems Change Your World Grant
1. Project Description
This project attempts to develop a teaching module, by utilizing Sun SPOT (Small Programmable Object Technologies) wireless motes and their built-in accelerometers, temperature sensors, and networks stacks, for the Embedded Systems Design course in the Electrical and Computer Engineering at Howard University. The course is designed to teach and practice wireless sensor network application design and development for knowledge aggregation and complex motion behavior understanding.
The current Embedded Systems Design course has been offered to upper level students of the departments of Electrical and Computer Engineering, and as a Technical Elective to Systems and Computer Science students. The course is centered on the development of stand-alone digital devices using medium-range microprocessors with assembly and C programming languages. The desired improvement of the course which is hoped to be realized by the proposed project is the combination of each digital device within a network so that a new system evolves from the collaboration of the devices, and the combined interactive and interconnectivity between devices, which improves the level of knowledge and understanding of the surrounding situations and behaviors. Each small device may not be aware of the entire situation, or be able to identify the object of interest. As each blind man in the fable of "the blind men and the elephant," the networked system of each individual devices would share the pieces of knowledge and, through an aggregation process yield an increase in the knowledge of each device and the system as a whole.
The ARM core-based Sun SPOT wireless sensors would make inroads for the Embedded Systems Design course to realize networked knowledge aggregation for object identification and complex motion/behavior understanding. For a foreign object identification, each on board SPOT would sense, measure, and respond to the object and then would provide information to a base station SPOT, which in turn combines them with a learning process. Further, a continued communication between each SPOT and the base SPOT would produce better information on the object. For complex motional behavior understanding case, multiple SPOT devices attached to the various places to recognize an object in motion would provide information processed from the accelerometer and send to a base SPOT, which would combine all motional information for an aggregated understanding of the object in motion.
The immediate benefit of the Sun SPOT course module would be the enhancement of students' understanding in a wireless sensor network architecture, communication stack and protocols, and hardware implementation of sensor technologies. Moreover, students would benefit being exposed to the advanced topics of swarm intelligence, knowledge sharing and aggregation, and motional and behavioral composition of a situation, which in all would produce competent minority engineers and scientists in this important area of emerging technology.
Project Outcomes and Deliverables:
The outcome of the project has two components: undergraduate learning outcomes and the research outcomes. The undergraduate learning outcomes are the major component of the project. By practicing the SPOTing "the blind men and the elephant," the undergraduate students of the course would achieve the following outcomes:
Even though the proposed project is mainly for undergraduate student education, the developed system would provide a good resource for research. The research outcomes include:
Deliverables include the Web posting of the project progress and the application note of completed application of object identification or motional behavior understanding.
Teaching Component and Student Participation:
The Sun SPOT course module would be adopted for the Embedded Systems Design course and would enhance students' understanding in a wireless sensor network architecture, communication stack, and protocols. In the class, students would experience all steps of hardware/software implementation of the SOPT based wireless sensor network. Teams of students would work together on the SPOTing "the blind men and the elephant" as their class project and they will program in Java, using a development tool such as SPOTWorld or NetBeans, the functions of the communication stack, accelerometer, data integration, and object/behavior identification. The results would be reported and demonstrated at the end of the class. It is expected that about a dozen students would be involved in the SPOTting project. The PI, as the instructor of the course, would guide the students through the basics of wireless sensors and networks, the software core of the Squawk virtual machine, Java development tool set-up, and SPOT's built-in features like the network stack and the accelerometer. A Ph.D. graduate student of PI, who finished his Master's thesis in wireless sensor network power management with a long experience of Java programming, would also help the students and the PI. Through the SPOTs and the SPOTing project, students would benefit being exposed to the advanced topics of swarm intelligence, knowledge sharing and aggregation, and motional and behavioral composition of a situation, which in all would produce competent minority engineers and scientists in this important area of emerging technology.
The PI and his graduate students have worked on the optimal power management of a single wireless sensor in a sensor network for use in object classification with approaches of Markov chain and matched filter to maximize the life span of wireless sensor in object classification sensor network. The work has been so far limited to models and Monte Carlo simulation. The provided SPOTs would provide a good opportunity to evaluate the model with physical wireless sensor for realistic power consumption and management in object identification. Further, the research can be extended to model development and verification for a system level power management for a whole entire sensors networked together for object identification.