Student project suggestions

Background

Rather than suggest particular projects, I much prefer to talk to people on an individual basis and try to find something that will be of mutual interest.

My current research interests are focussed around wireless sensor systems, pervasive systems (the IoT), control systems, robotics and autonomous systems, and security. However, I teach (or have taught) courses on control theory for robotics, wireless sensor systems, computer architecture using FPGAs, security, mobile systems, operating systems, networking, Linux internals, compilers, databases, distributed systems, etc.

In my group we build things, measure things, and analyse data. All my projects require two things - (i) a willingness to learn new things as necessary to finish the project, some of which might well be outside your comfort zone, and (ii) practical application, good testing and, in some cases, statistics on the data captured. In short, I enjoy supervising self-motivated people who want to do something useful and want to do it in a professional way.

I have almost no interest in web/database projects.

In general, in my group we build hardware for wireless sensor nodes or robots, we interface sensors to it, and we deploy this (i) on humans and animals, typically to measure behaviour and performance, and (ii) in both indoor and outdoor environments to measure anything from energy and water consumption to pollution levels. Once we have data, we undertake analysis on it and, where necessary, we invent new analytical techniques. The work involves building hardware, writing low level code to interface to sensors, data gathering, network protocols, security, and data analytics to name a few. The area has names that range from wireless sensor networking and pervasive computing to the Internet of Things. I have a separate interest in the use of technology in the education of schoolchildren.

Personal suggestions

For most of these, you need to be, or become, a member of the Institute of Making.

Developing world

1. Design, build, and evaluate a cheap networked weather station for use in the developing world. Landslides (which kill more people globally than do rivers and flooding) are triggered either by seismic activity or by prolonged and/or intense rainfall. Having the ability to monitor this is therefore crucial in determining when land may be about to slip. I would like to deploy a significant number of such monitors through a local NGO in Kalimpong in west Bengal where landslides are a major problem for the local population but where rainfall conditions vary considerably over distances of several km. We will do tests in the UK on the weather stations so developed, comparing the results to commercial products. The project will involve 3D printing, some simple hardware design and construction, and software development - including firmware, networking code, and data upload to the cloud.
2. Design, build and evaluate a cheap networked seismometer for deployment in the developing world. Many volcanic eruptions are from volcanoes that are considered 'dormant', which are therefore not monitored. However, it might be possible to detect the early signs of awakening through low-level seismic activity - but the seismometers must be sensitive enough to detect this and cheap enough to deploy in significant numbers since there are many dormant volcanoes. Such a seismometer would have uses in the context of other natural hazards, such as landslides. As above, this will involve 3D printing, some simple hardware design and construction, and software development - including firmware, networking code, and data upload to the cloud.
3. Design, build and evaluate a cheap and easily maintainable gas sensing rig for volcanoes that emit high levels of corrosive SO2 - this comes in upto three parts: ambient atmospheric gas, wearable gas sensors, and gas from the surface of the volcano itself.

Robots

Building

All build projects involve a design, build, evaluate process. In parts this will involve the simulation of robots before they are built, to test principles and to increase the scale of experimentation that can be done. For such simulations, the physical bot could be incorporated as hardware in the loop.

1. Create a robot for maintaining green walls - these are vertical structures full of plants that are becoming part of the green strategy of many urban workplaces. They have a tendency to dry out and need maintenance, which is tedious for humans, but could be easy for a robot that could move safely across the structure.
2. Create a cheap legged robot, (starting with one leg), in which at least some of the actuators are hydraulic and powered by a 3D printed microfluidic pump. Ideally, explore mechanisms for adapting leg compliance. This mimics the way that biology works in spiders, for example. The intention is to develop this idea for use in search and rescue missions. Yet again, this project will involve making, some simple hardware design and construction, and software development - including firmware, and control code.
3. Create a robot (a UAV or ground-based robot) that can autonomously deploy sensors AND retrieve a previously deployed sensor in an area of rough terrain. The housing of the sensor can be designed to make this easier, but this will involve making an appropriate gripper and using a mixture of GPS and machine vision to locate and collect the sensor.
4. Build a ground skimming UAV - this is intended to fly autonomously, primarily indoors (or on water), no more than ~5-15cm above ground, avoiding obstacles in the process. Assess how well it it possible to localise the robot using optical flow and SLAM. There are two options for this type of vehicle - a hovercraft (air cushion vehicle) or a WIG (wing in ground effect) vehicle. The control is non-trivial but likely to be interesting.
5. Build the electronics and control systems, from scratch, for a 3D printed UAV that was developed by a previous MSc student. The emphasis is on cost reduction and robustness since the envisaged deployment domain is on natural hazards in the developing world.
6. Simulate and then build very simple swarm robots, to be deployed in significant numbers. The evaluation process will likely involve hardware-in-the-loop simulation to provide appropriate scale.
7. Build a robot that gathers pond weed (or, possibly, plastics) from aquatic environments, that dries and then burns the recovered material as a means of providing an appropriate charge to a battery that powers the robot. A small solar panel can act to ensure that total battery depletion never occurs. In this way, the robot can be self-sustaining.
8. Build robots capable of navigating pipes oriented both horizontally and vertically - as a means of allowing robots to move throughout buildings via infrastructure built into the walls.
9. Build a flapping robot capable of autonomous flight and landing.
10. Build a tracked robot that is capable of autonomous navigation around unstructured rough terrain. This project involves both the construction of the robot and the processing of video to evaluate whether the robot is likely to be able to pass over a particular obstacle or whether it must navigate its way round it. A further extension would involve the use of maps created by a UAV in advance of (or even during) deployment. Ideally, such robots would have a battery that can be recharged from a small solar panel in extremis, to avoid their becoming permanently stuck in an inaccessible location.

Communication

1. Simulate and then deploy ground based robots or UAVs capable of creating and maintaining a sensor network by deploying small communication nodes and by acting as a data mule in a DTN-like scenario.
2. Create a control strategy for a UAV that bridges a communication link between two ground nodes.
3. Simulate optimised task allocation amongst robots in a swarm; specifically, this might be either DTN-like data collection from sensor nodes or search and rescue.

SLAM

1. Evaluate and adapt a SLAM/bundle adjustment system for localisation and 3D mapping of a steamy environment (such as one might find on a volcano).
2. Build a SLAM system from cheap radar components for use in smoky and dusty environments in which optical approaches are unsuited. Could start this with sonar, and move to radar.

Reinforcement learning

1. Many and various, generally involving learning to control a system. This could be learning low level controllers, planning, swarm coordination, etc. etc., and might include elements of both offline and online learning.
2. Dealing with failure - see e.g. ​https://arxiv.org/abs/1407.3501

Misc

1. Use a UAV to provide sound localisation of an audio source, specifically a whistle of the sort that many hikers carry in case of emergency.
2. Aside from audio localisation, I'd really like to be able to localise a radio source over several km, of the sort that could be generated by a mobile phone carried by a person who has become lost or injured. Ideally, this would be the GSM signal itself, but this is hard to access and so it would be possible to use a WiFi? or LoRa? signal in place of this, coupled with a directional antenna or multiple antennae on the UAV.

Disability

1. Explore geographic routing algorithms for the ARCCS (Accessible Routes from Crowdsourced Cloud Services) research project. The aim of this project is to explore how best to advise wheelchair-bound individuals about how to get from A to B, given information about fixed hazards, surface type, slopes, cambers and the particular disability. This is primarily a programming/algorithmics project, but has the potential significantly to impact the lives of disabled people. This work is expected to lead to a publication.
2. Work on indoor localisation with a wheelchair basketball team: they would like to be able to track their players, figuring out how far individuals have travelled and whether their tactics are working. This would use both IMU sensors developed for the ARCCS project and time-of-flight based localisation techniques. The same localisation techniques could be applied to other projects - tracking individuals with dementia in a museum environment; and tracking children playing in a forest school.
3. Develop an augmented reality version of wheelchair tennis, in which the ball is virtual and the physics of its flight can be adapted to allow for different skill levels or different levels of disability. There would be haptic feedback in the racket so that individuals got a sense of how they had hit the ball.

Enabling tech

1. Determine the increase in accuracy obtained by placing multiple cheap accelerometers on a single board. This is a piece of research that involves some prototyping, some programming, and the development of algorithms that fuse together multiple sensor inputs. A knowledge of linear algebra would be a significant advantage.

Collaborative suggestions

Music from environmental sensors (with Sound Matters)

Project Overview

Alongside our partners Commonland, AlVelAl? and Ecosystems Restoration Camps, Sound Matters is supporting efforts to restore degraded land in the Altiplano de Granada, Los Vélez and Alto Almanzora region of southern Spain. An essential aspect of this restoration work is the improvement of soil quality and, importantly, the communication to a broad audience the importance of soil fertility. Without healthy soil, a region cannot move from a state of languishing to flourishing.

Sound Matters aims to raise local awareness of the importance of healthy soil in landscape restoration initiatives by creating a Soil Composer. Using soil quality indicators (SQI) based on the physical, chemical, biological properties of soils, and other related environmental data of specific areas in the AlVelAl? region, one aim of this project is to create a musical interface for the soil. This interface will use a combination of soil analysis equipment, data processing, and music-generation software, to produce sounds and music which tell a unique story about the soil and show, through sound, how appropriate soil management techniques can create healthy soils that literally ‘sing’. This project melds the arts and sciences to create outcomes that are intended to affect behavioural change through raising awareness around the importance of measures to improve the health of soils - the basis of flourishing communities.

Partnership with UCL and student involvement

Sound Matters is proposing to work with staff and students from the Department of Computer Science at UCL to identify a set of key indicators of soil health and then develop an efficient soil quality indicator (SQI) sensor that can be used to:

• collect important information on soil quality that can be a source of data for ecologists working on the project alongside Sound Matters; and
• assist Sound Matters develop the Soil Composer concept.

In terms of the Soil Composer, data that will be collected by the SQI sensors will be used to create MIDI (Musical Signal Digital Interface) signals that can be used to affect, for example, notation, pitch and velocity and control signals for parameters such as volume, vibrato, panning, amongst others. The aim is to 'compose' music from soil data which will change in character as efforts in landscape restoration improve soil health. If possible, attempts will be made to live stream soil data so that people can listen to the composition in real time. In addition, a sound installation will be created which can be used to engage local people (especially farmers), living in the study area, with landscape restoration efforts and how these efforts can improve soil health and, by extension, the economic, environmental and cultural vitality of the AlVelAl? region.

Sensors for sand dunes

We would like to deploy a number of sensors in coastal areas as a pilot for a research proposal that we are creating with Birkbeck. These may well become buried by sand, and so need to be able to both sense and communicate in wet, sandy, environments. There are a range of possible projects, from 3D beach reconstruction from drone imagery (possibly augmented with cheap Lidar), through to tests of coding schemes and protocols for acoustic (or mixed acoustic/electromagnetic) communication, sensing strategies, and drone-based collection of sensor data. Come and talk to me if you're interested.