|robotics @ cyphynets|
A major research thrust at CYPHYNETS is theoretical and experimental research in various areas of robotics. We are self sufficient to develop our own robot platforms including mechanical design, mechatronics, control, perception, software and algorithms. We concentrate on developing unmanned ground vehicles for humanitarian applications such as landmine clearance and agriculture, in line with our lab's goal of developing cyber-physical systems for development. On the theory side, we employ geometric algorithms and systems theory to study fundamental problems in algorithmic robotics.
An overview of our recent activities in experimental robotics can be found at Winning the NI landmine detection competition: Lessons in sensors, systems and algorithms.
More details on specific projects can be obtained from the Publications page.
Also, check out the CMPE-633x series of graduate level robotics courses being taught at LUMS.
Marwa mine-detector ( مروہ)
Duldul all-terrain robot (دل دل)
Ha-Nun telepresence machine (ح ن)
iRobot Multi-robot coordination testbed (صفا)
ALVeDA (Autonomous Land Vehicle for Demining and Agriculture)
ALVeDA focuses on exploring the possibilities of using low cost sensor systems for real-time mapping, localization and navigation of an off-road mobile robot to be used in agriculture and mine-clearing purposes. More specifically, robust algorithms shall be developed for autonomous navigation in agriculture fields and demining in highly vegetated terrain. The focus of research collaboration is on key navigational capabilities related to obstacle detection, terrain classification and path planning, which are required for the operation of autonomous outdoor robots in rough terrain. Combining the Pakistani partners experience in humanitarian robotic demining with the knowledge-base of off-road field robotics in German side, robust and efficient solutions for outdoor robot navigation with low cost sensor systems in heavily vegetated terrain are expected.
Visual Servoing and SLAM Algorithms for Landmine Detection
The detection and removal of buried landmines is a global problem being faced by many countries. Due to the geographic extent of the problem as well as the danger associated with deploying a human into a mine-field, it makes sense to think about an automated, possibly robotic solution to the landmine clearing problem. Our motivation for developing a landmine detecting robot came from an international robotics competition by National Instruments (NI). In 2009, NI Arabia in collaboration with LebMAC (Lebanese Mine Action Center) invited proposals for the NI Mine Detection Robot Design (NI MDRD) contest in Lebanon. LUMS design named as Marwa went on to win the competition in 2011 with an award in the Vision & Sweeping Algorithms category. Our design focused on extensive use of vision algorithms, real-time embedded control systems and scalable software architectures to produce novel schmes in visual servoing of landmine detector arm and navigation via RGB-D based SLAM.
Spectral Methods for Motion Planning and Distributed Robotics
We explore the spectral properties of Probabilistic Roadmap (PRM) graphs and subsequently find the connection between spectral geometry of configuration spaces and the topological and geometrical obstructions encountered by motion planning algorithms. This can lead to new motion planning algorithms in robotics or perhaps a way of characterizing configuration spaces from a spectral perspective. Laplacian matrix gives important information about any graph. The eigen-vector decomposition of Laplacian matrix can be used to find algebraic connectivity of a graph. This is also called spectrum of graph. Every PRM graph has different spectrum, from where we get spectral properties. These properties can be used to detect narrow passages in expansive configuration spaces.
Testbed for Distributed Robotics