CAS::CSE Side Projects

Autonomous Insult Delivery

This project is designed as an indoor demonstrator, suitable for a platform such as the ActivMedia PeopleBot. It combines several of CAS's core research areas, including localisation, navigation, human-form recognition and human-robot interatction.

The aim of the project is to build a robot which can navigate a crowded room, say at a party or similar gathering. The need for strong navigation and obstacle avoidance skills is obvious. The robot should locate people, approach them, and then deliver an appropriately tailored insult, suitable to the person in question. So, for example, a short person might be addressed as "...", whereas an unually fat person might receive a remark such as "...". Follow up remarks such as could also be delivered.

It is imagined that this system will make an excellent demonstration for visiting dignitaries, government ministers, financial benefactors and the like. Indeed, the robots could also be trained to recognise specific individuals and create and deliver even more carefully personalised insults for these.

The problem of writing appropriate insults for all the various people encountered is an obvious stumbling-block for such a system. One suggestion is the behavioural-cloning techniques could be used, so that the robot could learn insults by observing suitable training examples provided by human experts. A ready source of such data might be obtained by taking the robot to Federal Parliament.

Feng Shui Robots

This is proposed as a key project in sensor desgin and multi-agent distributed data fusion. It will involve a number of different platforms, including small ground-based robots (such as the ActivMedia Pioneers) and autonomous air-vehicles as the DraganFlyer.

Feng-shui is well studied practive of influencing the "energy" or "chi" of an area by the appropriate arrangement of furniture and the placement of objects such as windchimes, potplants and water-waters around a home or office area, to promote positive energy.

However, in a dynamic working environment the day-to-day changes in office layout can play havoc with maintaining consistent positive chi. Enter the autonomous Feng-shui robots. A veritable army of small ground-based robots bearing pot plants and water features, plus autonomous air vehicles carrying wind-chimes can be distributed around the workplace. Each agent individually monitors the local chi level, and communicates with those robots around it to build a global chi-map. As the energy flow changes, the robots reorganise themselves autonomously to keep the positive energy at a maximum.

This project clearly has many important research issues involving sensing and multi-agent coordination: Can an effective chi sensor be designed? How should distributed chi sensors be integrated? Can a feedback system be designed, or learnt, which would maintain chi at a maximum? Experimental validation of such a system may also be an area of difficulty.


Rock-Paper-Scissors player

The aim of this project is to create an autonomous Rock-Paper-Scissors (RoShamBo) player capable of competing against a human player.

Rock-Paper-Scissors is a well-known, highly-competitive game for two players (see To summarise, players simultaneously nominate rock, paper or scissors with rock beating scissors, scissors beating paper and paper beating rock. The aim of the game is to win as many times as possible against the opponent by predicting their next move.

This project aims to combine a robotic arm and hand with a computer vision set up to enable play against a human opponent. Problems for the project include identifying when play starts (the prime), identifying the human opponent's move (rock, paper or scissors hand gesture), and formulating a strategy to beat the opponent.




Fashion Victim Identification

CAS is planning to compete in the forthcoming RoboCup Rescue competition. One component of this competition is the identifaction of victims in the environment. The Fashion Victim Identification project seeks to extend this work by developing algorithms and strategies for determining poorly dressed people.

The first part of this project entails gathering a suitable data set. The K-17 building at the UNSW campus provides the perfect environment for this task.

Key linkages include the Autonomous Insult Delivery project.