Integrating UMN Robots/Software with the ARL Infrastructure
The Army Research Laboratory (ARL) has the Asset Discovery and Control infrastructure. It is a flexible environment for organizing the different assets. The objective of this work is to integrate UMN platforms and software into this environment and expedite the technology transition.
- The UMN Scout is a miniature robotic platform that is ideal to be utilized in reconnaissance/surveillance missions.
- The Loper platform is a versatile robotics platform capable of operation in a number of environments. It has also been designed by UMN researchers to carry the UMN Scouts at the various locations when an urgent response is required.
More information is available at the project repository.
Effective Camera Placement for Human Activities Recognition Systems
Camera placement has an enormous impact on the performance of vision systems. As a result, this proposal focuses largely on the problem of task-specific camera placement. We attempt to determine how to place cameras relative to the motions of subjects of interest, in order to provide image input to a system so as to optimize that system's ability to achieve its task (recognize motions, take measurements, identify humans, etc.).
- Develop the best possible setup for human activity recognition.
More information is available at the project repository.
Scalable sWarms of Autonomous Robots and Sensors (SWARMS)
The SWARMS project brings together experts in artificial intelligence, control theory, robotics, systems engineering and biology with the goal of understanding swarming behaviors in nature and applications of biologically-inspired models of swarm behaviors to large networked groups of autonomous vehicles. Our main goal is to develop a framework and methodology for the analysis of swarming behavior in biology and the synthesis of bio-inspired swarming behavior for engineered systems.
- Develop novel systems-theoretic concepts to bridge discrete and continuous mathematics to synthesize controllers and estimators to perform abstracted collective behaviors
- Enable control and monitoring of large groups of robot agents and sensors with a minimum number of human operators
- Look for bio-inspired solutions not mimicked behaviors
More information is available at the project repository.
Dynamic Task Allocation and Vehicle Mission Planning in Constrained Environments
Recent developments of autonomous unmanned vehicles and sensor technology have motivated extensive applications in mobile surveillance. There is a strong demand at the software level to manage the whole system with few operator interferences. The specific goal of this project is to develop efficient decentralized algorithms for dynamic task allocation and optimal motion planning for scenarios involving multiple heterogeneous ground, air, and/or sea-based autonomous vehicles. The immediate focus is on:
- Develop abstract models of both tasks and vehicles so that mission allocation and high fidelity motion planning can be accomplished in a computationally efficient manner
- Develop Anytime algorithms for maximum flexibility in meeting mission requirements
More information is available at the project repository.
Mapping of Large Scale Indoor and Outdoor Environments for Fast Spatial Awareness
Robot localization without any map knowledge can be effectively solved with combinations of expensive GPS and IMU sensors if we can assume that GPS cannot be jammed and works well in urban canyons. Given accurate robot poses from GPS/IMU, one can quickly establish a quasi-dense 3D map of the environment if provided with a full 2.5D laser scanning system and if the system can detect moving objects around. This is how vehicles, in their majority, navigate autonomously in recent DARPA Challenges. GPS-denied environments as well as indoor spaces render GPS based algorithms unusable. We propose a system for Monocular Simultaneous Localization and Mapping (Mono-SLAM) relying solely on video input. Our algorithm makes it possible to precisely estimate the camera trajectory without relying on any motion model.
More information is available at the project repository.
Autonomous Landing Site Selection and Confirmation for Unmanned Helicopters
Autonomous unmanned helicopters could be employed for a variety of tasks such as: surveillance and mapping of urban environments, search-and-rescue operations in hazardous environments (e.g., radioactive sites), border control, scouting hostile areas, surveying national forests for emergencies such as fire, and gathering massive statistical data for various studies. One of the challenges of the autonomous use of unmanned helicopters is autonomous selection of sites that can be used for easy landing with minimal risk of failure. An unmanned helicopter constructs an elevation map using prior information and sensor data during the flight. An autonomous landing site selection and confirmation system needs to analyze this elevation map in order to either select an obstacle-free landing site that is safe for landing or direct the helicopter to obtain more sensory data to confirm the safety of the possible landing sites. This problem of autonomous selection and confirmation of landing site can be cast as a planning under uncertainty problem. The uncertainty is in the status of the landing sites, whereas the action space is the set of possible commands for the helicopter: where to go, what landing site to sense using its on-board sensors, and finally where to land.
More information is available at the project repository.
Place Recognition for Fast Rescue Operations
Mapping technology has recently been enriched with ground-borne imagery provided by the main mapping services like Google Maps/Earth or Microsoft Virtual Earth. These are panoramic images tagged with geographic coordinates of the associated viewpoint. Assume now a new photo at a completely different time, weather, and illumination taken by a different and less expensive camera with lower resolution and dynamic range. Moreover, imagine a rescue scenario in an environment partially destroyed. How can a victim convey her/his position by sending an image with a camera phone? The challenge is in the different imaging conditions but also in the fact that only a small part of the picture might exist in the database. We propose an algorithm that depends both in the similarity of appearance as well as in the geometric/structural consistency between two given photos.
More information is available at the project repository.
Autonomous Urban Patrolling and Surveillance
Our goal is to build fast, reliable, safe and autonomous vehicles that will revolutionize transportation systems in urban environments. Our approach is to leverage state-of-the-art advances in sensing, control theory, machine learning, automotive technology and artificial intelligence to build robotic car. We already have a significant level of experience in this technology as a finalist in the DARPA Urban Grand Challenge in 2007.
More information is available at the project repository.
Surveillance and Environmental Monitoring with Smart Camera Networks
This project is dedicated to the design and development of a smart camera and sensor network. We will develop the hardware and algorithmic foundation for a network of cameras and heterogeneous sensors to form a smart network capable of providing the user with high level situational awareness of the environment covered by the individual component sensors. The network will then be capable of identification and tracking of user specified objects of interest, fusing and correlation of data acquired by others sensors with imagery from the camera network, and other global conclusions that can be drawn through integration of the network.
More information is available at the project repository.
Indoor Micro-Flyer
Design and build an easily controllable hovering rotorcraft; Develop mechanisms and control algorithms that outperform existing flyers in size, agility, ease of control, and/or cost/maintenance and to synthesize control algorithms for non-holonomic attitude control.
More information is available at the project repository.