Skip to main content
Menu

Oxford Robotics Institute | Team ORIon - The Team

LEADERSHIP TEAM

Our Leadership Team sets the vision of the team and supports team members to achieve their goals and get the most out of participation, through social events and training.

TASK-LEVEL PLANNING SUB-TEAM

The Task-Level Planning sub-team is responsible for developing and maintaining high-level state-machines to implement autonomous behaviours to accomplish tasks, drawing on capabilities provided by the other sub-teams.

PERCEPTION SUB-TEAM

The Perception sub-team is responsible for robust object detection and classification. The sub-team implements state-of-the-art machine learning algorithms to train and deploy predictors and classifiers, using high-dimensional LIDAR and RGB-D point cloud data from the HSR.

MANIPULATION SUB-TEAM

The Manipulation sub-team is responsible for developing robust manipulation actions, including grasp-pose planning, grasp motion planning, and pick-and-place behaviours.

Manipulation in action @Robocup 2019 "take out the trash task"

SEMANTIC MAPPING SUB-TEAM

The Semantic Mapping sub-team is responsible for developing semantic mapping and reasoning capabilities, as well as knowledge representation capabilities.

HUMAN ROBOT INTERACTION SUB-TEAM

The Human Robot Interaction sub-team is responsible for human speech recognition and human-machine-interaction. They develop user-friendly interfaces (e.g., audio, gesture) to permit non-technical users to command and interact with the HSR, as well as solve problems such as multiple speaker recognition and speech direction identification. They are also developing an “ethical black box” system to help end-users, developers, and regulators understand what the robot is thinking and doing with better transparency, and thus increase trust and encourage the adoption of robotics in society.

NAVIGATION SUB-TEAM

The Navigation Sub team is responsible for robot motion planning. Currently Navigation actions are controlled via the ROS Action server provided by the tmc move base ROS package. Previously, we used a topological map (i.e., discrete locations connected by directed edges). Edges corresponded to navigation actions the robot could perform to transition between locations. We are currently exploring incorporation of topological navigation into the current navigation system to permit location-based action selection (e.g., checking if a door is open).