E-commerce is a thriving industry and as the demand for online delivery keeps increasing, so too do high costs, energy expenditure, carbon emissions and congestion associated with it. Many of these negative effects are a result of the so-called "last mile", the final stretch of the delivery process. Hence the need for extra capacity and efficiency in this final stage of the delivery process.

Autonomous delivery robots offer a solution as they are energy efficient and less polluting. Additionally, a robot can be reduced in size and operate on sidewalks or pedestrian walkways, contributing to less congestion on roads. A robot can also delivery at any time outside normal working hours.

Developing such a system brings attention to some of the key questions in robotics research, such as perception, outdoor navigation and manipulation, all of which still have many open research questions. One particular area which has had little research focus is the perception system of a mobile robot when manipulating objects in its environment.

Aim:
The overall aim of the PhD is to develop a robust perception system for a mobile robot to successfully manipulate objects (mobile manipulation), which could be used in applications such as last-mile delivery.

Objectives:
1. Review the state-of-the-art in perception systems used for manipulators, mobile robots and mobile manipulators.
2. Outline the specification for the perception system based on the requirements for mobile manipulation for last-mile delivery.
3. Develop and test the perception system for manipulation.
4. Develop and test the perception system for mobility.
5. Integrate the two perception subsystems into a single perception solution for mobile manipulation.
6. Test and analyze the overall perception system in a simulation environment and then in the real-world. This could involve some mock-up test on a delivery problem for example, driving up to a house and delivering a package through a letter box.

Research Questions:
Sensing: What are the most appropriate sensing technologies for a mobile manipulator for last-mile delivery? For example, can the robot rely on just a camera? Or will it require depth information or additional sensor modalities such as lidar or sonar?
Perception: How does the robot gain an understanding of its environment in order to achieve its goal? For example, how does it detect relevant objects in a cluttered space?
Grasping and Manipulation: How does the robot choose the best place to grasp an object? How does it grasp objects varying in size, shape and texture?
Mobile Manipulation: How do you develop perception systems to allow for both manipulation and mobility?