Student Projects

Autonomous Vehicle Series

Autonomous Vehicle Series is a set of projects concerned with autonomous vehicle (AV); being a car, a drone or any moving autonomous systems. These project aims to create a scalable, safe, and efficient multiagent framework for managing autonomous vehicles at intersections.

Intelligent vehicle technology is progressing very rapidly and recent advances suggest that autonomous vehicle navigation will be possible in the near future. At modern-day intersections, traffic lights and stop signs assist human drivers in conducting their vehicles safely through the cross traffic. However, in the future, with computers "behind the wheel", will it make sense to have intersection control mechanisms that were designed with today's human drivers in mind? With all the advantages computerized drivers offer - more precise control, better sensors, and quicker reaction times - it is believed automobile travel can be made not only safer and easier, but much more efficient.

The main aim of this series is to have a set of simulation tools for application of AV in real life context. All projects are based on the AIM4 simulator implemented at the University of Texas. The code with selected publication can be found at http://www.cs.utexas.edu/~aim/.

 

Due to the complexity of the AIM4 API (in Java), PRMI-PRMC students must have achieved at least 80% in the JAPRC module assessment to take on a project from this series.

 

MagicRoundAbout lblot-16
Autonomous Vehicles intersection Management: A case for the roundabout. PRBX-PRIA-PRIY-PRMC-PRMI Available

This project is part of the Autonomous Vehicle Series. Please read the description of the series given above.

In this project we want to extend the model developed at the University of Texas to include roundabouts. In addition we want to evaluate the efficiency (or lack of) of roundabouts compared to usual intersection in an AV only model.

ThatsMySpot lblot-19
Autonomous Vehicle Parking Management. PRBX-PRIA-PRIY-PRMC-PRMI Available

This project is part of the Autonomous Vehicle Series. Please read the description of the series given above.

Autonomously-driven cars are only a few years away from becoming a common feature on our roads. These self-driven vehicles hold the potential to significantly change urban transportation. One of the most important changes will not happen during the trip from origin to destination, but rather when these vehicles arrive at their destinations. An autonomous vehicle will leave its passengers at their destination and will then park by itself, waiting to be called to pick them up later on. This behaviour will have important implications on door-to-door trip time, traffic congestion and parking costs.

Ferreira et al. [1] proposed a new parking concept where the mobility of parked vehicles is managed by a parking lot controller to create space for cars entering or exiting the parking lot, in a collaborative manner. They have shown that the space needed to park such vehicles can be reduced to half the space needed with conventional parking lot designs. They also showed that the total travelled distance of vehicles in this new parking lot paradigm can be 30% less than in conventional parking lots. The aim of this project is to extend the research done by Ferreira etal. at the University of Porto. In their model, they have only considered vehicles of similar size, using slot of fixed size. In addition, in their work they did not take into account the behaviour pattern of a users, for example a casual shopper (staying for an hour or two) versus a commutter (9 to 5 weekdays).

In this project we will attempt to address this shortfalls. In addition, their model is based on a typical car park, we would like to explore car park targeted to specific business needs, such as car park for car manufacturer, retailer, and cinema complex. In addition you will be looking at gathering a dataset to validate your model.

You will have to implement your solution in Java using a simulator based on the AIM simulator, which has been further developped at the UoY.

 

[1] M. Ferreira et al., "Self-automated parking lots for autonomous vehicles based on vehicular ad hoc networking," 2014 IEEE Intelligent Vehicles Symposium Proceedings, Dearborn, MI, 2014, pp. 472-479. doi: 10.1109/IVS.2014.6856561

WheresMySpot lblot-20
Intelligent Road for Parking Management. PRBX-PRIA-PRIY-PRMC-PRMI Available

Intelligent Road Panel For Parking Management

 

This project is part of the Autonomous Vehicle Series. Please read the description of the series given above.

Autonomously-driven cars are only a few years away from becoming a common feature on our roads. These self-driven vehicles hold the potential to significantly change urban transportation. One of the most important changes will not happen during the trip from origin to destination, but rather when these vehicles arrive at their destinations. An autonomous vehicle will leave its passengers at their destination and will then park by itself, waiting to be called to pick them up later on. This behaviour will have important implications on door-to-door trip time, traffic congestion and parking costs.

In the mean time, the road will be shared between self-driving cars and human-driven vehicles. The aim of this project is to devise new algorithm to improve car park capacity using intelligent road panel [1] in a mixed environment of autonomous and non-autonomous vehicle. The car park should be able to guide a driver to a given spot, whilste managing autonomous vehicle via V2I communication. In addition, the carpark should be able to change its layout depending on the type of vehicle (size, autnomous or not, handicap parking permit, and so on) arriving to the carpark.

You will have to implement your solution in Java using a simulator based on the AIM simulator, which has been further developped at the UoY.

 

[1] http://www.solarroadways.com/

  LETsChange   lblot-21
Intelligent Road Self-Layout Management. PRBX-PRIA-PRIY-PRMC-PRMI Available

Solar road

 

This project is part of the Autonomous Vehicle Series. Please read the description of the series given above.

The aim of the project is to devise a collaborative model between the modular panel of intelligent road [1] to change the layout of the road according to the traffic. We could start with a three lanes road, where the two lanes closest to the pavements have traffics in opposit direction, and the middle lane change traffics direction depending on vehicle density in a given direction of travel. The change of direction should be dynamic and done in real time.

One of the benefit of such application is to provide fast middle lane access to emergency vehicle, or facilitate overtaking slow moving vehicle localy.

You will have to implement your solution in Java using a simulator based on the AIM simulator, which has been further developped at the UoY.

 

[1] http://www.solarroadways.com/

Roundandround lblot-22
Autonomous Vehicles intersection Management: A case for the Giratoire Interchange. PRBX-PRIA-PRIY-PRMC-PRMI Available

 

Giratoire interchange

This project is part of the Autonomous Vehicle Series. Please read the description of the series given above.

In this project we want to extend the model developed at the University of Texas to include raised roundabout interchange. In addition we want to evaluate the efficiency (or lack of) of such interchange model compared to usual intersection in an AV only model.

MagicRoundAbout lblot-23
Autonomous Vehicles intersection Management: A case for the Clover Leaf Interchange. PRBX-PRIA-PRIY-PRMC-PRMI Available

Clover leaf interchange

 

This project is part of the Autonomous Vehicle Series. Please read the description of the series given above.

In this project we want to extend the model developed at the University of Texas to include Clover Leaf type interchange. In addition we want to evaluate the efficiency (or lack of) of this kind of interchange compared to other intersection in an AV only model.

DiamondsR4ever lblot-24
Autonomous Vehicles intersection Management: A case for the Diamond Interchange. PRBX-PRIA-PRIY-PRMC-PRMI Available

Diamond interchange

 

This project is part of the Autonomous Vehicle Series. Please read the description of the series given above.

In this project we want to extend the model developed at the University of Texas to include roundabouts. In addition we want to evaluate the efficiency (or lack of) of roundabouts compared to usual intersection in an AV only model.

Other Projects

The selection of project below are not part of any series. They are not related to each other and cover a more broader area of research.

   SeedsofLove     lblot-26
AI for the Oware Game: A Monte Carlo Tree Search approach. PRBX-PRIA-PRIY Available

Oware Game Board

 

The game of Oware is a variant of the game of Awari, one of the oldest known strategy games. It is believed to have originated in Ethiopia about 3500 years ago and has since spread across Africa. Oware is played with a hollowed out plank of wood and a number of stones or seeds, the number varies depending on the variant of the game (see image above).

From a Computer Science point of view, Oware is a game of perfect information, otherwise known as combinatorial games. These games are classified as two-player games, with no hidden information, no chance moves, a restricted outcome (win, lose and draw) and with each player moving alternately.

Some works have been done on developing AI for Oware such as Galilea (Decision Making System for the Game Oware . Home University: Facultad de Informática de Madrid (Spain) & Institut für Algorithmen und Kognitive Systeme (IAKS), Fakultät für Informatik der Universität Karlsruhe (Germany) March 23, 2004.), and also some variants of the game such as Gifford et al. (2008) for Mancala, and Davis and Kendall (2002) for Awari.

 

The project aim is to develop a Monte Carlo Tree Search algorithm for the game of Oware, and compare its performance against other AI, and if possible human players.

Student Defined Projects

The following projects have been defined by students, and therefore are already allocated to them. These are not available to other students. Please choose a project from the selection above.

   TheFuzz     lblot-15
Is smart fuzzing worth implementing? PRBX Not available, Allocated to Raluca Morel

Motivation:

Sophos Group Plc (Sophos hereafter) is a security software company which also provides internet security solution, which includes antivirus, encryption and data protection, web filter. As part of the file scanning process, the antivirus engine must recognise and parse the content of files in order to discover malware. Since it is impractical to conduct manual testing at a very large scale, fuzz testing is one of the techniques used by Sophos. It is a cost-effective means which transmits input data to the target program to trigger anomalous program behaviour and increase the code coverage. If the data is not randomly generated and follows certain key patterns, the redundant cases do not get tested, which makes the fuzzing be more efficient, “smart”.

 

However easy and fruitful fuzzing may seem, there are certain costs associated with this method, especially when it comes to raw computational power. Reaching a new path through the code can become more expensive and time-consuming than moving to subsequent input files and potentially reaching it there. Moreover, programs run twenty to thirty times slower inside the Valgrind MemCheck tool than outside of it [1] and discovering a new path does not always lead to discovering a new vulnerability. On top of that, smart fuzzing has additional costs: understanding the appropriate degree of “smartness” that should be implemented in order to not make it too restrictive or loose; the implementation and the integration with the existing fuzzing systems.

 

Aim:

The aim of the paper is to outline the costs and benefits of smart file parsers fuzzers over the blind ones. If time allows, an investigation of the optimal fuzzing time using a file, given some attributes (i.e. length), will also be conducted.

 

In order to achieve that, background investigation will be needed in order to understand the vulnerabilities that can arise from a loose file format parser and how they can be exploited. For instance, the PDF file header which identifies the file as PDF, can be used to avoid being correctly scanned by an AV product. According to the specification, the first line of a PDF file shall consist of 5 characters %PDF- and a version number [2]. However, Adobe Reader can load and parse files which do not necessarily contain those 5 characters, therefore it raises an issue for the AV products that only check for the first few bytes as their means of recognising a PDF file.

 

Given the scope of the project, the paper will focus on a particular file format.

   Liar-Liar     lblot-25
Making AI for a board/card game. PRMI Not available, Allocated to Chris Hughes

Motivation:

For this project, I intend to create a piece of software that allows a user to play a game against one or more AI players. While the game is susceptible to change, the two options I am currently considering are Liars Dice and Scopa.

In order to create an AI player, I intend to research algorithms, such as the Monte Carlo Tree Search and minimax search, and examine how they can be applied in this case. Ideally, I will be able to create my own algorithms, but I am also willing to adapt existing algorithms to my purposes if this proves too great a task; providing a full justification in this case.

Past Projects

ThatsMySpot lblot-13
Autonomous Vehicle Parking Management. PRBX Rebecca Milligan (2016-17)

This project is part of the Autonomous Vehicle Series. Please read the description of the series given above.

This project is concerned with the management of parking spaces for Autonomous Vehicles (AV). The context of the project is relatively futuristic, where only autonomous vehicles are allowed in the car park. In this context, we assume that the car park's centralised system has the necessary access-right to control all AVs entering the car park once the user leave the AV at the entrance of the parking.

The aim of the project is to extend the simulator developed at the University of Texas to include Car Park management. You will have to develop an algorithm to manage the car park, and test its suitability and performance. Measure of performance could be "waiting time" of the user when collecting their AV, energy used by the AV to park.

Another interesting challenge is when all parking spaces are occupied. Additional AVs could be accepted if they keep moving around the car park (entering a loop) until a spot is available. For "usage fairness", where all AVs should share the cost of energy used in the loop, you will have to consider a kind of round robin where an AV releases its spot to another AV from the loop, and in turn enter the loop itself.

  LETsMerge   lblot-14
Autonomous Vehicles Lane Merging Management. PRBX Callum Hewitt (2016-17)

This project is part of the Autonomous Vehicle Series. Please read the description of the series given above.

In this project we want to investigate Centralised vs Decentralised approaches (Note: PRMI-PRMC student will focus only on the centralised approach).

The centralised approach will have a main agent situated in the area of the merging lanes (for example secondary road joining a motorway). Its role is to allocate which vehicule will be able to access a given lane at a given time. The aim is to reduce the amount of traffic jam, while reducing the total waiting time of vehicle wanting to join the motorway.

While this solution might be satisfactory for merging lane between two motorways, it is not suitable for the case of a road incident closing one or more lanes. This may happen at any point on the motorway and a central agent may not be close enough to manage the traffic. A decentralised approach, where vehicle close to the incident may take control of the traffic would be more suitable. Therefore the second part of the project is to devise such a model. This could be based on the election protocol for multi-agents.

  QuizzFlix   lblot-18
Building user profile for Movie Streaming using Games. Sultan Seidalin (2015-16) Thesis (pdf)

Building user data via Gamification. Information is money, and accessing user preference is difficult. In 2014, Netflix removed its public API, leaving many app developper without access to user preference in the domain of video streaming. In this project, we will be investigating if the use of movie based quizz game could help us develop a user profile for video streaming.

You will need to devise and develop a mobile app (on the platform of your choice), in order to gather data about a user, and from that data develop a user movie preferences profile. The mobile app need to be in the form of a game that could be shared and played by multiple users.

You will need to design an experiment to evaluate the validity of your approach.