Post Graduate / PhD Projects
In this project, a network-based urban pedestrian traffic light control problem is firstly formulated as a scheduling problem, aiming to reduce the total waiting time over a time horizon. Firstly, a novel mathematical model consisting of several logic constraints is proposed to describe the pedestrian flow in the urban traffic network and its dynamics in each junction is based on the cell transmission model but also involves the characteristics of the conventional two-way crossing (north-south and east-west) of the crosswalk topology in each intersection. Besides the individual pedestrian model, we also integrate our pedestrian model with the vehicle network model to formulate a linear optimization problem by developing a single objective function with different weightfactors on each side, aiming to investigate the impact of pedestrians on vehicle delays when simultaneously providing pedestrians with convenience.
We develop ultra-wideband impulse radio (UWB-IR) based cooperative and distributed positioning for unmanned platforms. The ultra-wide bandwidth in UWB results in well separated direct path from the multi-paths enabling more accurate ranging using time of arrival (TOA) of the direct path in cluttered environment. The peer to peer range between unmanned platforms can be derived from time difference of arrival (TDOA) of signals transmitted simultaneously from multiple vehicles. The range of information can then be exchanged among UGVs for collaborative positioning.
In this project, we address urban traffic signal control in a scheduling framework, where the dynamics of an urban traffic network controlled by traffic lights is described by a novel model, which inserts mixed logical constraints into a cell transmission flow dynamic model, capable of capturing the nonlinear relationship between each outgoing link flow rate and the corresponding upstream and downstream link capacities and the past traffic light signals. With a control goal of minimizing the total network-wise delay time, we translate the traffic signal control problem into a centralized mixed integer linear programming problem solvable by several existing tools. To overcome the potentially high complexity involved in the centralized approach, we propose a distributed computational architecture based on Lagrange multiplier method and subgradient method, which allows the computing intelligent methodologies to be implemented. Some under-going researches include the modelling of heterogeneous traffic system and implementation on VISSIM simulation platform.
In this project, urban traffic control is implemented in a scheduling framework. A Vehicle On-board Dynamic Route Planning Module is considered as an additional feature to this framework. A User-oriented Route Guidance approach has been developed, by extending Dijkstra’s Algorithm to dynamically integrate speed profile functions to derive shortest travel times. At the moment, we calculate the optimal paths for given speed profile functions over a short prediction horizon. Speed profile functions are currently derived from the predicted network traffic state resulting from the optimal traffic signal control framework implementation. Our aim is to simultaneously optimize traffic signal settings and calculate shortest paths, possibly through use of multi-objective optimization.
Post Graduate / Masters Projects
The main aim is to develop a novel platoon management protocol on the simulation platform, for basic platoon maneuvers such as split and merge. It includes developing a communication protocol for vehicles in the platoon to exchange beacon messages for performing the maneuvers. The channel switching (between the control (CCH) and service channels (SCH)) is a safety feature that is added to combat any accidents or mishaps from happening. It also includes broadcast simulation where the non-platooned vehicles are also communicating in the surrounding.
This work focus on the development of a novel PID controller for vehicular platooning of heavy duty vehicles on road networks of Singapore. The PID controller helps the follower vehicles to track the speed of the leader vehicle with a pre-fixed distance between them. The work also entails communication constraints and their effect on platooning.
In current Wireless vehicular communications and network, DSRC/WAVE is the only wireless technology that can potentially meet the extremely short latency requirement for road safety messaging and control. The messages are send on control channel (CCH) and service channels (SCHs). What this project focus on is the effect in platoon management when vehicle receiver keep alternating the CCH and SCH. The whole communication and platoon protocols has been simulated in NS3 and MATLAB and has been used to track communication delay and packet drops. Moreover, platoon merging and splitting for range determination has also been simulated as a part of this work.
Providing high quality video streaming over multi-hop V2V (vehicle to vehicle) networks is a challenging problem because wireless channels are noisy, error prone and time varying. A multi-hop network coding technique, called Batched Sparse (BATS) Code, is able to recover packet errors and improve the delivery rate of streaming video. However, the traditional BATS code was designed for the wired internet, not for fast varying wireless channels. In this project, the student will work with our researchers in the NTU-NXP Smart Mobility Testbed Project to adapt the BATS code for video streaming over multi-hop V2V networks. Commercial V2V modems will be used and real-time multi-hop video streaming will be tested on the road. The student will be required to do C++ and Python programming.
With the high development of artificial intelligence, machine learning and pattern recognition are playing an increasingly important role in object detection and recognition from images or video sequences. This technique can help people to analyse and extract significant information of the image or video more efficiently and accurately. This dissertation conducts an experiment on vehicle detections from campus surveillance video sequences with Deep Learning Method.
Computer vision study based on deep learning algorithm provide relatively precise result in frame and video detection. Among all kinks of deep learning frameworks and algorithms, Caffe and faster-rcnn perform outstanding in both detection speed and accuracy rate. In this dissertation, Caffe and faster-rcnn are applied to detect the pedestrian in campus monitor based on VGG network and ZF network. To improve the detection mAP, the network is trained using VOC 2007, VOC2012, VOC 2007+VOC 2012 and the superposition of other labeled pedestrian datasets.
Person Re-Identification is the new technique compared to the person identification, which is more difficult and complex. And it is easily affected by the surrounding environment and person’s gestures. Although Person Re-ID has made great progress theses years, there are still lots of problems which need to be solved. In this project, the basic knowledge about person identification was introduced and reviewed. And then the image based Person Re-ID and video based Person Re-ID are introduced. Besides, the Top-push Distance Learning (TDL) model was introduced and studied to apply to the experiment.
Vehicular WiFi (or car-to-car WiFi) enables real-time information exchange (such as location, speed, steering direction, braking status, cyclists/pedestrians presence on the road) between car and car, car and motor bike/bicycle, or car and pedestrian. This enables the vehicles to have better awareness of their surroundings, thus improving all road users' safety.
For example, adjacent cars can use vehicular WiFi signal to exchange GPS locations so that if 2 cars who are in each other's blind spots get too close, safety warning messages can be flashed to the 2 drivers to avert collision Different vehicles may use different types of Human Machine Interface (HMI) to display the safety enhancement messages. In this project, the HMI for passenger cars with a 7" display, and the HMI for motorcycles with a 1.4" smart watch or tablet display, will be developed using Android programming and tested on the road using the Smart Mobility Testbed facility installed in NTU.
The IEEE 802.11p standard is a new car-to-car WiFi technology that enable cars to communicate directly with cars to share GPS locations and other safety-related information. The objective of this project is to use the 802.11p modem and access point equipment made by NXP and Cohda Wireless to measure and analyze the communication channel quality when the channel is busy with many active users. The channel quality metrics to be measured include SNR, throughput, channel busy ratio etc.
The smart electro-mobility project involves electric-assisted bicycles. The e-bikes are equipped with smart device capable of tracking the vehicle’s position, energy level, condition, sending vital information to a central hub, and authentication. The e-bikes will also communicate with the vehicular communication network infrastructure deployed by the smart mobility test bed (SMTB) in School of EEE and specially-equipped cars moving on the roads.
The student responsibilities include:
1. Study the communication network (IEEE 802.11p DSRC) within the campus and develop a compatibility strategy for the e-bikes smart device.
2. Establish communication between the e-bikes and other entities of the SMTB.
3. Testing of the communication quality for smart mobility applications
This project will be carried out in collaboration with the company PSA Peugeot-Citroën Research (the second largest car manufacturer in Europe).
With video surveillance systems becoming more prominent, there have been numerous research studies carried out on detection and tracking on a subject of interest. This study aims to detect and track vehicles from video using a static camera mounted on a street lamppost using algorithms implemented on MATLAB. Four videos were executed to assess the robustness and stability of the system. As a typical video streams for about 30 frames per second, all the videos were set to 29 frames per second at a resolution of 640x340. The program would output two video players, namely video frames and foreground mask respectively.
While GPS can provide accurate location estimates in an open environment, an accurate GPS fix occurs only if the device has line of sight to at least four GPS satellites. In an urban environment with high rise buildings, ground vehicles typically do not have direct lines of sight to so many GPS satellites; therefore GPS localization becomes intermittent or even corrupted in some cases. In this project, we use DSRC signals and other sensor signals to aid in vehicular localization. Experiments will be conducted under various parameter settings and scenarios in order to evaluate the effectiveness of our localization approach.
Accidents happen when drivers do not follow the traffic rules well. For example, while driving, whenever a driver wants to change lane, be it to the left or right, the driver has to use the signal lever as an indicator to tell other road users his/her intention, However, due to lack of awareness and laziness, some drivers occasionally do not signal their intentions, and that is when and why an accident happens.
A Driver Assistant System will be developed here. There are sensors installed at both side mirrors of a car as well as at the front and the back of the car. The sensors at the side mirrors will sense any oncoming vehicles on the adjacent lanes beside the car at a defined distance. However, it will be limited to a top speed of 50 km/h. When the driver signals his/her intention to change lane, the system will check whether it is safe for the driver to change lane based on a set of algorithms, and then send out a voice message to advise the driver accordingly.
In addition, the sensor at the front of the car will be used as an indicator if the car gets too close to another car ahead while the sensor at the back of the car will indicate that the driver needs to speed up slightly. Likewise, the system will, based on a set of algorithms, send out a voice message to advise the driver to slow down or speed up accordingly. Other features will be explored and implemented in the system.
Traffic accidents cause loss of lives. Some causes of traffic accidents include tailgating, abrupt change of lane and failure to check blind spot. To enforce the safety of drivers and motorcyclists, a Vehicle Smart Alarm System will be developed, where this system will send different alarm sounds depending on the various potential dangers encountered on the road.
For example, when another car is coming up on a driver's blind spot, the distance sensor of the system, be it the left rear or right rear one, will sense that the other car is too near, and the system will then alert the driver. Another example is where two cars are changing into the same lane from the left and right, which may cause an accident. These two cars will be oblivious to each other until they are dangerously close, which will be too 700late for them to avoid collision with each other. The distance sensors at the right and left side of a car can help the driver by alerting him/her (via the system) in advance that another car on the right or left is too close to his/her car.
This project will require the development of both the Hardware & Software for this system. And this project has obtained a Merit Award at the MHA SAFE Competition 2015.
Selected Student Assignment Reports based on a 3-hour lecture by Prof Lu Yilong on Intelligent Transportation Systems (ITS)