Projects:2018s1-160 UAV Platform for Cognitive AI Agent

From Projects
Revision as of 17:07, 21 October 2018 by A1660069 (talk | contribs)
Jump to: navigation, search

Students

Junyi Jiang

Zhi Cao


Supervisors

Prof. Michael Liebelt

Mr. Xin Yuan

Introduction

This project aims to use Street Agent to develop an suitable experimental platform which is able to reproduce the Ant's foraging behaviours. This project has developed an entire ant foraging logical by using Street language and an multi-agent interface system for Street Agent. Own designed platform has been proved the Street Engine can be used to reproduce simple cognitive behaviours of ants.

Background

In biology, ants walk randomly and leave a trail of pheromone when they foraging, once an ant find food, ant can follow pheromone trail back to the nest, other ants will follow this pheromone to approach food when they find this pheromone trail. Street Cognitive Architecture is first developed based on Soar Cognitive Architecture at The University of Adelaide. It is aimed be used for artificial intelligent(AGI) applications. The street language has developed for building street engine, which is based on pattern-matching rules. In 2018, honour project team at the university has developed own designed street agent logical and an multi-agent entire suitable experimental platform using street agent to achieve ants cooperation foraging behaviours, including recognize food, nest location and pheromone. During project period, the street language has been used to create an artificial general intelligent system.

Motivation

Street Engine is a new developed computer architecture which is developed by The University of Adelaide. In previous years, two groups has been proved Street Engine is able to achieve advanced cognitive computation in real-time, such as Tic Tac Toe group. But the Artificial General Intelligence agent cannot be limited in one or two single function achievement.

A suitable experimental system for Artificial General Intelligent Street Agents remains as a challenge to a the Street project research group. Therefore, Our group expect to build a virtual and physical platform with embedded Street Agent to achieve multi-agent with multi-functions to reproduce simple cognitive behaviours of insects, such as ants foraging behaviours.

Technical Background

Street

Pciture.PNG

As shown in figure above, street includes Street Engine and Street language. The street language is used on the street engine directly. Therefore, the street processing has reduced the latency and power consumption. Street engine uses global memory to operate every single processing element. Street Engine is core system of the Street Agent, The Street Language is parallel structure which means the system does not force the system following flow control, every single command in system is independent. comparing current architectures, the real time processing efficiency has been improved.

V-rep simulation

Virtual robot experimentation platform was used in our project. It is involved to test our Street Agent logical and test our project design. The remote control interface is used to communicate designed Street Agent with virtual rovers.

Arduino Uno R3 with WIFI

Physical robot platform was built by using Arduino board with necessary sensors. Arduino board and 3D printed rover structure. Arduino rover is used to collect the environment information and receive the movement command which is from Street agent, such as avoiding boundary, approaching food source, approaching nest location and following pheromone trail.

Interface System Design

All system need to be controlled via own deigned Street agent logical. For all information from environment need to be translate and send to the agent, agent will send the decision back to platform. The flexible processing module and action module have been established for ensure the interaction of agent and platform. File:Picture.PNG