Ongoing Research at MRL

We are investigating methods for generating coverage patterns (via waypoints) of environments with obstacles. We use the Boustrophedon cellular decomposition (BCD) algorithm to break the world into cells and critical points, and then solve the Chinese Postman Problem (CPP) to generate an optimal traversal through these cells. We are deploying and validating our algorithm both using a simulated robotic platform as well as using an unmanned aerial vehicle.

We are developing an interface for controlling mobile robots that combines aspects of graphical trajectory specification and state space programming. Our objective is to facilitate the programming and debugging of robot execution plans, containing both scheduled trajectories and activities, as well as contingency plans and failsafe behaviors.

We examine the construction of a vision-based interface for human-robot interaction and control for autonomous robots in arbitrary environments. The goal is to enable a human operator to control and program the robot without the need for any complicated input interface, and also enable the robot to learn about its environment and the operator. We investigate the applicability of machine learning methods - supervised learning in particular -- to train the vision system using stored training data. Combining our novel algorithms for visual tracking, visual detection of human motion, gesture recognition, visual languages and a quantitative evaluation model, we aim to create a robust HRI scheme using vision as the interaction modality. Towards implementing this framweork, a substantial amount of systems infrastructure development has also been undertaken.

We are interested in understanding how mobile robots can learn to interpret information arising from physical contact with the world. In particular, we want mobile robots to learn to distinguish between terrain type (for a legged amphibious robot) or indoor surfaces (for a generic wheeled robot using a sensitive tactile feeler). Our method is based on a novel unsupervised learning technique that trains a classifier by exploiting the naturally occuring spatial or temporal continuities.

This project deals with coordination and intereaction between different types of robot vehcile, for example between our air vehciles and our underwater systems. One of the main themes of this work is to explore methods for efficient yet highly distributed task partitioning.

This project deals with underwater and amphibious robotics, and is the umbrella for a host of different specific research probjects at McGill, as well as in collaboration with our associates at York University. Most of the sub-projects enatil work with out Aqua family of underwater hexapod robots.

We are developing tools and techniques for inferring environment structure from a network of widely separated range sensors.

We are currently working on inferring complete range maps where only video images and very limited range data is available. We allow a mobile robot to rapidly collect a set of video images and very limited amount of range data from the environment and then infer the rest of the map it does not capture directly. Our goal is to facilitate the building of 3D environment models by exploiting the fact that both video imaging and limited range sensing are ubiquitous readily-available technologies while complete volume scanning is prohibitive on most mobile platforms.

Appearance based recognition using Principal Components Analysis with the added ability to account for varying backgrounds. This is done using an attention operator to focus on the object to be recognised and performing PCA only on the sub-windows within the object.

Several members of the mobile robotics group are assembling components of our software infrastructure into a real-time mobile robotics testbed.

geometry applet, robotics applet

This project deals with the exploration of an unknown environment using two or more robots working together. Key aspects of the problems are coordination, and particularly rendezvous, between the robots, and efficient decomposition of the exploration task.

more information

This project involves shape modelling based on a combination of local curvature information at multiple scale, and global superquadric surface fitting.

abstract

We are examining techniques for the creation and management of virtual reality analogues for the real world. This includes the automatic acquisition of image-based VR images, as well as the automated selection of viewpoints and scenes of interest. Further information on the image acquisition system is available by following the link.

more information

Methods for learning, encoding, detecting, and using visual landmarks for mobile robot pose estimation.

Autonomous navigation using sensory information often depends on a usable map of the environment. This work deals with the automatic creation of such a maps by an autonomous agent and the minimal requirements such a map must satisfy in order to be useful. One aspect of this work is the analysis of how uncertainty either in the map or in sensing devices relates to the reliability and cost of navigation and and path planning. Another aspect is the development of sensing strategies and behaviours that facilitate reliable self-location and map construction.