Schedule

Sensor fusion is a ubiquitous yet ever-challenging task in robotics. The challenge becomes even more critical under the constraints of aerial robots, which have limited payload and computational capabilities. In this talk, we will discuss the theoretical foundation for fusing information from different sensing modalities, namely visual, inertial, ranging, lidar, and 3D prior sources. The theory is put into practice through multiple real-world applications conducted by our group over the years. We will also address some prevalent challenges that are still awaiting innovation from the community. 

This talk presents a novel framework for developing a formation of nonholonomic mobile vehicles that takes into account communication constraints and task requirements. The framework is based on a cost function that includes both task-related and communication constraint-related costs, as well as a new concept of compatible task and communication constraints. The talk demonstrates that, if the task and communication constraints are compatible, then a family of formation control laws can be developed through off-the-shelf optimization algorithms to solve the formation problem. The talk also shows that gradient-like control laws can be utilized to find non-unique local optimal solutions of the optimization problem and achieve the desired formation for the group of mobile vehicles. An illustrative example is presented to demonstrate the effectiveness of the proposed framework in achieving the desired formation while ensuring collision avoidance, even when communication is limited to visual camera sensors 

Significant recent advances in unmanned aerial vehicles calls for convenient platforms to support experiments and demonstrations. Unmanned aerial vehicles such as quad-rotors and multi-copters have become popular for this purpose. However, the indoor usage of these unmanned aerial vehicles (UAVs) is limited by flight duration per battery charge (typically less than 20 minutes) and safety concerns to humans sharing the same physical space. Safety nets or cages provide protection, but sacrifice the potential for human robot interaction experiments. We develop the Georgia Tech Miniature Autonomous Blimp (GT-MAB) as flying vehicles for indoor experiments that support safe -interaction between human and robot swarm. The GT-MAB has relatively long flight duration up to two hours per battery charge. Furthermore, the blimps are naturally cushioned and do not cause any pain when collide with human. It offers a fun experience that often encourage physical contacts with humans. We have developed vision-based feedback control laws that enable the GT-MAB to detect and track humans. We will report recent progress made on the modeling and control of the GT-MAB. Advanced control techniques achieve more stable flights and more agile motion that enable new applications for the GT-MAB. 

Robust multi-agent SLAM is a key technology for efficiently autonomous exploration of large-scale complex environments. We propose a robust multi-agent SLAM system that combines LiDAR, visual, and inertial sensors, which can work in challenging environments such as low texture, poor illumination, or/and heavily dynamic scenarios. Our system adopts a novel sensor fusion strategy that effectively integrates data from LiDAR, visual, and inertial sensors in pose estimation and loop closure detection. By leveraging their respective strengths, multiple agents can collaboratively work together to achieve high-precision, real￾time localization, and construct a globally consistent environmental map. The proposed system has potential applications in various fields, such as autonomous driving, scene reconstruction, and post-disaster rescue. 

As the perception, planning, and coordination technology going mature, autonomous robots are given higher expectations by people. Aerial robots, and their swarms, now are required not only to fly out of laboratory, but also to finish more complicated tasks. To this end, building smarter drones with sophisticated functionalities where perception and planning modules are coordinated or even coupled is an attractive research topic. In this talk, I will introduce some new methods of aerial robots developed in my group, by which we may broaden the application range and intelligence level of drones. Then, based on real-world requirements, some systematic solutions for specific tasks are presented, where the architecture, algorithms, engineering considerations, as well as closed-loop performances, are explained. Finally, I will turn to some of our most recent research, on which we are working towards a perception-planning coupled, flexibly coordinated, and large-scale aerial swarm system. 

In robotics, we often design the hardware separately from the controller. This divide-and-conquer approach is efficient, but often leads to overall inefficiencies. In this talk, I will present some work focusing on tightly coupled hardware & control co-design for aerial robotics systems. Specific examples presented will include a system designed to exploit angular momentum to increase robustness to disturbances, and two separate vehicles that change their shape to improve their capabilities. Such shape-shifting robots borrow ideas from natural systems, and exploit internal degrees of freedom to achieve novel capabilities. 

As the development of mobile robots matures, there is an increasing amount of interest in expanding the functionality of such robots through developing multimodal locomotion. As compared to land–water or land–air hybrids, the design of air-water vehicles is much less straightforward due to the fact that both mediums are three-dimensional fluid spaces and there is inherent disparity in fluid properties between them. As such, the development of these vehicles has received limited attention until very recently. Nevertheless, the potential applications of such vehicles range widely from surveillance, oceanic data collection to heterogeneous robot team operation. In this talk, we aim to present the hardware design, modeling and control of an aerial-aquatic vehicles are capable to operate in the air and underwater with special consideration for propulsion and thruster configuration to effectively cope with the vast differences in the medium properties of water and air.

This talk addresses the coverage control problem of a network of heterogeneous mobile sensors, where the goal is to minimize a coverage cost function which is defined to be the largest arrival time from the mobile sensor network to the points on a circle. A necessary and sufficient condition for the global minimization of the coverage cost function is firstly derived. Then, distributed coverage control schemes with input saturation are developed to drive the sensors to the optimal configuration such that the necessary and sufficient condition is satisfied. Under the distributed coverage control schemes, the mobile sensors' spatial order on the circle is preserved throughout the network's evolution and thus collision between mobile sensors is avoided. Finally, simulation results are presented to illustrate the effectiveness of the proposed distributed control schemes. 

In this lecture, we aim to present a fully autonomous and fully functional infrastructure inspection and information management system with advanced AI and multiple unmanned aerial systems (UAS) technologies. The system includes sophisticated unmanned aerial hardware platforms and software systems for automatic flight control, task and motion planning, artificial intelligent algorithms and software platform for image and infrared data processing, i.e., crack, spalling, delamination and other defect detections, and building information modeling (BIM) and management system integrated with detailed geographical information systems (GIS) and digital twin (DT) technologies. Compared with the manual inspection, the system that we have developed has the advantages of being more economical, safer, flexible and efficient. It can also be adopted for other industrial applications, including smart traffic management and smart cities.